Navigating the complexity of AI adoption in psychotherapy by identifying key facilitators and barriers

Navigating the complexity of AI adoption in psychotherapy by identifying key facilitators and barriers | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Navigating the complexity of AI adoption in psychotherapy by identifying key facilitators and barriers Julia Cecil, Insa Schaffernak, Danae Evangelou, Eva Lermer, Susanne Gaube, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6546781/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 14 You are reading this latest preprint version Abstract Artificial intelligence (AI) technologies in mental healthcare offer promising opportunities to reduce therapists’ burden and enhance healthcare delivery, yet adoption remains challenging. This study identified key facilitators and barriers to AI adoption in mental healthcare, precisely psychotherapy, by conducting six online focus groups with patients and therapists, using a semi-structured guide based on the NASSS (Nonadoption, Abandonment, Scale-up, Spread, and Sustainability) framework. Data from N = 32 participants were analyzed using a combined deductive and inductive thematic analysis. Across the seven NASSS domains, 36 categories emerged. Sixteen categories were identified as factors facilitating adoption, including useful technology elements, the customization to user needs, and cost coverage. Eleven categories were perceived as barriers to adoption, encompassing the lack of human contact, resource constraints, and AI dependency. Further nine, such as therapeutic approach and institutional differences, acted as both facilitators and barriers depending on the context. Our findings highlight the complexity of AI adoption in mental healthcare and emphasize the importance of addressing barriers early in the development of AI technologies. Health sciences/Health care Scientific community and society/Social sciences Scientific community and society/Social sciences/Psychology Mental healthcare Artificial intelligence Technology Adoption Figures Figure 1 Figure 2 Introduction The treatment gap in mental healthcare, resulting from a discrepancy between the need for treatment and its availability, remains a significant global challenge 1 , 2 . One in eight people is likely to develop a mental disorder during their lifetime 3 . However, with a global median of only 13 mental health professionals per 100,000 people, each would theoretically be responsible for over 7,000 individuals, including approximately 960 requiring care 3 . This shortage is reflected in long waiting times, for example, in Germany, patients wait an average of 14.5 weeks for an initial appointment 4 , with other countries showing similar patterns 5 . This unrealistic demand for and shortage of mental health professionals, often coupled with inadequate coordination among psychotherapy providers, hinders optimal care 2 , 6 . Several additional factors contribute to inadequate treatment, including access barriers (e.g., travel costs), fear of stigmatization, and the misinterpretation of mental health symptoms as normal responses to social and economic challenges, rather than as treatable conditions 2 . The advent of digital health technologies enabled by artificial intelligence (AI) presents new opportunities to help bridge the treatment gap and transform mental healthcare. AI-enabled technologies include tools for screening, diagnosing, predicting, and treating mental health conditions, as well as solutions designed to enhance therapists' performance and streamline administrative tasks 7 – 10 . While technologies for screening, diagnosing, feedback and practice management primarily focus on decision support and efficiency improvements, AI-enabled treatment applications actively engage both patients and therapists in the psychotherapeutic process. This paper focuses on the latter category of applications within the psychotherapeutic treatment process, emphasizing their dynamic interaction with both user groups – a feature that distinguishes them from other application areas. These technologies leverage advanced algorithms to customize treatment 11 . For instance, they can support therapists’ work by providing psychoeducation or therapeutic interventions specifically tailored to individual patient’s needs 12 . Oftentimes, these AI-enabled treatment tools focus on depression and anxiety 8 , 11 . For instance, Natural Language Processing (NLP) models can assist in selecting the most effective therapeutic approach for treating depression, such as cognitive behavioral therapy or drug treatment, tailored to the individual patient 13 . Overall, AI-enabled technologies have the potential to enhance the therapists’ work experience by reducing assessment times 14 , while providing improved and more timely support for patients 14 , 15 . AI adoption in psychotherapy Despite these possibilities, both patients and therapists demonstrate reservations about these technologies. Mental health professionals show a low intention to use AI-enabled technologies, especially patient-centered tools, including those used for treatment 10 , 16 . Among patients, research has demonstrated that dropout rates for AI-enabled psychoeducation applications, for example, range from 5–34% 17 . This reluctance might highlight a critical issue: much of the existing research on technology adoption in psychotherapy focuses on the design and functionality of AI applications rather than addressing user needs 18 – 23 . Understanding user needs and perceptions towards AI technologies is crucial to identify, address, and ideally prevent barriers that could hinder appropriate technology adoption. Theoretical framework for understanding barriers to AI adoption Recognizing that many healthcare innovations fail due to a variety of factors, the Nonadoption, Abandonment, and Challenges to the Scale-up, Spread, and Sustainability (NASSS) framework was developed by Greenhalgh et al. 24 . It provides a structural approach to analyze multiple dimensions and complex dynamics shaping the use of AI in healthcare by capturing barriers and facilitators to technology adoption in seven domains (see Fig. 1 ). In mental healthcare, the NASSS framework has been used to guide adoption research on video therapy consultations 25 , and internet-delivered cognitive behavioral therapy (CBT), and CBT in virtual reality 26 . The first domain, condition(s) or illness , focuses on the nature and complexity of health conditions targeted by the technology in question 24 . It further encompasses clinical and sociocultural aspects, such as comorbidities 24 . Research on the adoption of AI technologies designed for mental disorders, such as depression or eating disorders, has mostly focused on already developed specific applications 18 , 23 . However, little research takes a bottom-up approach by first exploring which disorders users consider suitable or not for AI treatment tools. The second domain, technology or technologies , refers to the technology’s characteristics, including their features, functionality, and usability 24 . An established requirement for AI adoption in mental healthcare is the technology’s foundation in robust scientific evidence and its capacity for personalization to meet patient's specific needs 19 , 27 . Interestingly, specific features vary in relevance. For instance, tracking and feedback are generally viewed positively, while features like goal setting and relaxation audios are less preferred and may pose barriers to adoption 28 . In the next domain, value proposition , concerns and benefits, such as the perceived desirability, efficiency, and cost-effectiveness of the technology, are examined 24 . From an economic perspective, some applications, such as AI-enabled computer assisted therapy, have been shown to be cost-efficient 29 . However, the use of digital technologies for the diagnosis and treatment of depression, does not seem to lead to healthcare cost savings 30 . The fourth domain, adopter system , includes users’ perceptions of adoption and workload considerations, addressing factors such as training needs and individual-level barriers to adoption. On the therapist side, these issues include insufficient training, both before and during the adoption of AI technologies 27 , 31 , 32 . Additionally, there are concerns about the dehumanization of care and perceived power imbalances between therapists and patients due to shifting dynamics in the triadic relationship of therapists, patients, and technology 27 . From the patient's perspective, key barriers also include a perceived loss of human connection, and lack of professional oversight 33 , 34 . Next, the domain organization(s) , addresses the readiness and capacity of an organization for a specific technology 24 . Although limited research addresses barriers at this level in mental healthcare, financial challenges hindering AI adoption were identified in other healthcare contexts 35 , 36 . Specifically, a lack of funding, costs associated with expert input for labeling and curating AI training data, and technology maintenance create significant barriers to adoption at the organizational level 35 , 36 . The sixth domain, wider context , considers the structure, dynamics, and capacity of the healthcare system in which the innovation is introduced 24 . The domain examines how the new technology interacts with existing practices, policies, and resources. Prior research on AI in psychotherapy identified obstacles such as cost coverage 33 and regulatory issues, specifically the absence of structured implementation frameworks 27 , 31 , 32 . The final domain, e mbedding and adaptation over time , addresses the challenges and strategies associated with expanding the reach and impact of the innovation beyond initial pilot or research settings 24 . This domain has received limited attention in research 35 , 37 ; however, some factors such as ongoing technology maintenance seem to play a role 36 , 38 . Despite a growing body of research on AI applications in mental healthcare, a gap remains in understanding the adoption barriers from a bottom-up perspective, particularly given the limited findings in this specific field of healthcare. This study, therefore, investigates the needs and barriers from both patient and therapist perspectives through the NASSS framework, addressing areas typically explored through a top-down lens, such as organizational and system-level structures. In doing so, we aim to generate insights that can inform best practices for stakeholders, including developers, helping to proactively address user reservations prior to the development of AI technologies. Methods Study design This qualitative study used focus groups to explore the needs of both patients and therapists for AI-enabled technologies in psychotherapy. Focus groups provide an opportunity for an exploratively in-depth analysis of an area with limited prior research and enable nuanced, contextual understanding of participants’ perspectives and underlying reasoning 39 . This pre-registered study ( https://osf.io/nzb2k/?view_only=c9fcc933d2e8421c981d73cd740f95fd ) received ethical approval from the [blinded for review], identifier: 2023-11-V-144-R. It adheres to the 32-item Consolidated Criteria for Reporting Qualitative Research guidelines 40 , see supplementary material on OSF. Recruitment and Participants The study included two participant groups: therapists and patients. Therapists were recruited via emails distributed among psychotherapy training institutions and patients were recruited via flyers in and distribution channels of two German universities and two German patient organizations. Therapists were required to be licensed or undergoing psychotherapy training. Patients were required to be enrolled in psychotherapy and to be diagnosed with either depression or anxiety disorder. These conditions were chosen due to their high global prevalence and burden 41 . A total of 32 participants were recruited, comprising 19 therapists and 13 patients. Participant characteristics can be found in Table 1. Each focus group consisted of four to six participants, with therapists and patients placed in separate groups. Using homogenous groups enhances the external validity of the research as it allows for in-depth discussions between participants, providing each group with space tailored to their specific experiences and perspectives 42 . Additionally, this approach helps mitigate any potential power imbalances between therapists and patients, creating a more open and balanced discussion environment 42 . Table 1 Sample demographics Patients (n = 13) Therapists (n = 19) N (%) or M (SD) N (%) or M (SD) Age 30.23 (10.05) 33.32 (10.55) Gender Female 10 (76.9%) 15 (79.0%) Male 2 (15.4%) 4 (21.1%) Non-binary/ third gender 1 (7.4%) Mental disorder Depression 10 (55.7%) Anxiety 5 (27.8%) Other 3 (16.7%) Profession Therapist in training 17 (89.5%) Therapist 2 (10.5%) Therapeutic approach (Cognitive) behavioral 12 (57.14) Psychodynamic 6 (28.57) Psychoanalytic 2 (9.52) Systemic 1 (4.76) Workplace Practice 3 (10.34) Private practice 1 (3.45) Specialist hospital for psychiatry, psychotherapy, psychosomatic medicine or neurology 7 (24.18) Rehabilitation clinic 4 (13.79) (University) outpatient clinic 10 (34.48) Community mental health center/ counseling center 1 (3.45) Other 3 (10.34) Professional experience (in years) 3.89 (4.78) Note . Combined age for therapists and patients: M all = 32.06, SD all = 10.3 Data collection Data was collected between March and May 2024. The online focus groups were conducted in German and via Zoom (Zoom Video Communications, Inc.), and lasted 58.36 minutes on average ( SD = 2.05, min = 56.23, max = 61.49). Participants signed a data protection and consent form prior to their participation. Additionally, they received a general description of AI and AI-enabled technologies in psychotherapy to facilitate their familiarity with the subject in advance. The description was intended to offer a general overview of the various tasks AI can perform in psychotherapy, rather than concentrating solely on a specific tool. The description can be found in the supplementary material on OSF. At the beginning of each focus group, participants were informed about the purpose and procedure of the discussion and that their participation was voluntary and could be withdrawn at any time. The focus groups proceeded with a brief recap of AI-enabled technologies in mental healthcare to reinforce participants’ understanding. The focus groups were facilitated by two moderators. One moderator guided the discussions, while the other posted the discussion guide questions in the chat and addressed participants’ questions. For the questions, a semi-structured discussion guide was used. The discussion guide questions are based on the NASSS framework 24 and related studies in this field (for instance, see 26 ). Questions and prompts only differed in the wording to address each groups’ distinct characteristics, see online Supplementary Information. Data Analysis The sessions were recorded and transcribed verbatim using Trint (Trint Limited). Content analysis was conducted with the transcripts using MAXQDA (VERBI Software). Data analysis followed the deductive and inductive approach outlined by Fereday & Muir-Cochrane 43 . First, a code template was deductively developed based on the theoretical structure of the NASSS framework, which provided seven broad coding domains 43 . The template contains the names of the domains with a definition and how to apply them to the transcripts. Second, to ensure the reliability of the codes, two researchers (JC and DE) familiarized themselves with the transcripts and third, the transcripts were coded independently according to the respective seven domains of the NASSS. Overall, complete sentences were used as the smallest coding unit and multiple codes were applied to the same sections if necessary. Subsequently, an inductive approach was employed to identify emerging themes from the data, which were organized within these predefined NASSS domains. In this fourth step, two researchers coded the transcripts independently again. It is worth noting that, during this step, the original NASSS subcategories were coded when explicitly mentioned in the focus groups. However, participants predominantly generated new categories that exceeded existing NASSS subcategories. As a result, we opted not to constrain new categories to the predefined NASSS subcategories, ensuring that the analysis remained true to the participants’ contributions. Fifth, as new themes emerged, they were grouped together and assigned concise labels. Lastly, after the independent coding, both researchers compared their categories and the names of the categories were refined and merged, resulting in a final codebook. Both researchers then independently coded all transcripts again. The intercoder agreement after this step, measured as a Cohen κ value of 0.73 for segments with at least 90% overlap, indicates a moderate to substantial level of agreement 44 . Any discrepancies in coding during this step were discussed with a third researcher (IS) and resolved through consensus. Data availability The discussion guide, COREQ checklist and the description of AI and AI-enabled technologies in psychotherapy given to participants prior to the study is available on OSF ( https://osf.io/9rt52/?view_only=a1a4b6a50d8f4d62b3b685eee6651a9b ). The codebook and exemplary codes supplementing the study results are available in the online Supplementary Information. Results The final codebook contained 561 codes, ranging from 83 to 102 codes per focus group. The seven domains of the NASSS framework with the emerging subcategories and descriptions are summarized in Table 2. Each category of the NASSS was classified as a facilitator or barrier when at least two-thirds of the codes reflected a positive or negative influence on AI technology adoption, respectively. In cases when less than two-thirds of the codes reflected an overall positive or negative influence, the category was classified as mixed . The number of codes for each category and their valence (i.e., barrier, facilitator or mixed) can be seen in Fig. 2 . Table 2 The NASSS domains, categories, and descriptions regarding the adoption of AI technologies in psychotherapy NASSS Domain Category Description Condition The nature and complexity of mental health conditions targeted by AI technologies. Type of mental health condition The specific mental health conditions that AI may target. Severity of mental health condition The range of illness severity that AI may address. Technology The characteristics of AI technologies. Useful technology elements The specific elements that are considered useful for AI. Customization to users’ needs The extent to which AI can be tailored to specific users' needs. Supplementary use The use of AI as an adjunct rather than a replacement for traditional therapy. Usability for multiple user groups The extent to which AI is suited for use across different groups. User-friendliness The ease and intuitiveness with which users can use AI. Evidence-based effectiveness Evidence supporting AI’s effectiveness in improving mental health outcomes. Value proposition The value of AI technologies for users. Increasing availability of access The ability for users to access AI at their convenience. Improved efficiency Potential improvements in efficiency offered by AI. Shifts in depths of therapy elements The risk of AI oversimplifying therapy and reducing nuanced insights. AI as third element The role of AI as an additional agent during psychotherapy sessions. (Commercial) conflict of interest The influence of commercial interests in the adoption of AI. Adopters User factors influencing the adoption of AI technologies. Affinity for technology Users’ comfort level and enthusiasm in using AI. Training requirement The need for training to use AI effectively. Lack of control The potential loss of control over therapy processes due to AI. Trust and openness The degree of trust in and openness for AI among users. Technology understanding The knowledge or technical literacy required for effective use of AI. Lack of human contact The absence of social and emotional connection after the integration of AI. Table 2 (continued). NASSS Domain Category Description Changes in the importance of conventional psychotherapy The potential of AI changing the perceived need for conventional therapy. Technology dependency (Over-) Reliance on AI technologies. Persistence of safety-seeking/ avoidance behavior The potential of AI to reinforce users' avoidance of difficult interactions or situations. Organizations The readiness and capacity of an organization for AI technologies. Differences in the type of institution Underlying differences in the type of therapy-providing institutions influencing the use of AI. Differences in therapy methods The compatibility of different therapeutic approaches with the potential use of AI. Digitalization The extent of digital infrastructure present in institutions. Pre-use testing opportunities The chance to try the AI at least once before use. Other incentives Existing services or support structures necessary for adopting AI. Resource and workload considerations The resources required for integrating AI. Wider system The structure, dynamics, and capacity of the system in which AI technologies are introduced. Cost coverage The extent to which the use of AI is covered by insurance or other funding sources. Economic feasibility Economic considerations regarding the adoption of AI. Regulations and regulatory environment Legal and policy frameworks guiding the use of AI. Data protection measures Measures ensuring data protection for the use of AI. Liability Determining accountability for AI outcomes. Transformations of existing structures Shifts in the structures of the mental healthcare system. Embedding and adaptation over time The adjustments in the adoption of AI technologies over time. Speed of implementation process The time it takes to integrate and adopt AI. Ongoing scientific validation The necessity of ongoing research to validate the efficacy of AI prior to its use. Note Numbers on the upper right corner of each category indicate the total number of codes assigned to that category. Tile size reflects the frequency of codes. 1. Condition For the domain condition , two categories emerged as mixed factors (labeled M). 1.1. Type of mental health condition (M) : Participants identified 15 mental health conditions that could potentially be addressed by AI technologies, with anxiety disorders and depression being the most frequently mentioned. Participants generally agreed on the potential usefulness of AI technologies for the following conditions: panic disorder, sleep disorder, hyperkinetic disorders, conduct disorders, substance use disorders, gender dysphoria, and borderline personality disorder. For instance, one patient explained: “If there's […] a button: 'I'm having a panic attack. Help me', and it tells me to do different exercises or whatever, I would take it.” (PA01). Opinions were more divided for anxiety disorders, depression, eating disorders, ADHD/ADD, and suicide. For instance, therapists highlighted both potential benefits and risks of AI-based data tracking for eating disorders. Tracking physical activity and identifying deviations from usual activity patterns could provide valuable insights, as patients might not be fully aware of behaviors like excessive walking, which could contribute to difficulties in weight gain. However, such tracking could become a new obsession for patients, leading to an unhealthy focus on monitoring behaviors. Participants were skeptical about using AI technologies for conditions such as social phobia, psychosis, and paranoia. For instance, one therapist remarked that patients with paranoia or psychosis might be even more reluctant to adopt AI technologies, as these tools could amplify their existing suspicion and mistrust of their surroundings, leading to heightened discomfort and anxiety. 1.2. Severity of mental health conditions (M) : Participants emphasized the need to distinguish between mild, moderate, and severe mental health conditions to ensure appropriate adoption of AI. AI was unanimously deemed useful for milder conditions, but participants were hesitant about its application in more severe cases. They suggested a cautious, stepwise approach, starting with mild conditions and gradually expanding to more severe conditions. 2. Technology In the technology domain, five categories were perceived as facilitators (labeled F) and one as mixed. 2.1. Useful technology elements (F) : Participants had various ideas for beneficial technology elements that would facilitate the adoption of AI technologies. Table 3 presents an overview of the designated technology features. Table 4 Useful technology elements retrieved from the focus group discussions Feature Information purposes Psychoeducation General information Emotional and behavioral management Emotion regulation Mood tracking Mood detection Skills a Relaxation exercises Logging and monitoring/ tracking Triggers Different types of diaries Activity tracking Recording of critical events Identifying patterns a Therapist support and process analysis Data analysis for therapists a Analysis of the therapy process Monitoring the development of the therapeutic relationship Decision support for treatment options Decision support for the therapeutic approach Self-management and organization Homework assignments Reminders Creating a daily structure (Self-help) Exercises Emergency plan Reinforcement plan/ system Safety and risk management Detection of suicidality Technical and interactive features Avatar/ representation of the therapist Automatic application filling Voicemail Note . a = not specified further The most prominent features included psychoeducational content, mood tracking, different types of diaries, and self-help exercises like imaginary journeys. For instance, the value of mood tracking beyond fixed points was highlighted, especially when daily routines are disrupted, to provide deeper insights into a patient’s emotional state. 2.2. Customization to users’ needs (F) : Another critical factor for the adoption of AI technologies in mental healthcare is its ability to be customized to users’ individual needs. Personalization might help alleviate symptoms, while its absence might even worsen them: “Every panic attack manifests itself differently, and what helps one person may not help another at all. You tell one person to concentrate on your breathing, and then they start to hyperventilate. You tell the other person to concentrate on your breathing and they calm down. […]” (PA03). 2.3. Supplementary use (F) : Further, participants from both groups stressed using AI only as a supplement technology, not as a replacement for traditional therapy or a standalone solution, since they view human therapists as irreplaceable in the therapeutic process. 2.4. User-friendliness (F) : Participants stressed that the technology should be easy to use, especially to reduce cognitive and emotional strain for individuals who may already feel overwhelmed. “I think it should be very intuitive and not so complex, because especially if you have depression or something like that, you just don't have the strength to get into it.” (PA08). 2.5. Evidence-based effectiveness (F) : Finally, participants emphasized the need for evidence supporting the effectiveness of AI technologies in improving outcomes. “There should be studies that prove that AI-supported interventions have an effect, which is why it makes sense to use it, but the effect is not as great as psychotherapy from a real person.” (TH14). 2.6. Usability for different user groups (M) : Another AI adoption factor was its usability for multiple user groups. Three different sociodemographic areas were mentioned in which adoption barriers should be addressed: age, socioeconomic status, and culture and language: First, older generations should be enabled to use emerging technologies. “It is important to have a different perspective for children and young people than for adults or geriatric patients. I can possibly see difficulties there […], due to their probably rather limited use of digital devices. In other words, I ask myself, are these older patients then excluded? That should, of course, be avoided.” (TH18). Second, usage should be possible regardless of personal financial background, as participants emphasize potential inequalities for those with a lower socio-economic status. And third, adaptations to different cultural and language backgrounds should be considered. “[…] I believe that there will also be a large section of the population that you do not deal with because it will then be very complicated to translate into simple language applications or actually offer something in their native language.” (TH11). 3. Value proposition In the domain value proposition , five categories emerged as either facilitators or barriers (labeled B) to AI adoption in mental healthcare. 3.1. Increasing availability of access (F) : A key benefit of AI technologies is accessibility, both in terms of location and timing. Participants highlighted the advantage of using AI flexibly in different places and its critical role in timely supporting patients during moments of crisis by providing immediate access to therapy resources. “If I imagine: ‘Okay, I can fall back on previous therapy content, I'm in a crisis at the moment and can simply call it up’, I don't necessarily need the therapist for that, that in turn can also have positive effects in terms of autonomy.” (TH18). 3.2 Improved efficiency (F) : AI could enhance efficiency by automating certain tasks, such as data analysis and administrative processes, enabling therapists to focus on patient care and reducing the strain caused by high patient volumes and long waitlists. 3.3 Shifts in depths of therapy elements (B) : Concerns were expressed that AI technologies might oversimplify complex therapeutic dynamics and reduce the value of nuanced observation. AI might overlook interpersonal dynamics and subtle behavioral cues, which are essential for forming accurate therapeutic judgments. “[…] I wonder if […] [AI is] too simplistic because we [humans] are actually much more complex than that, more than could be broken down into an algorithm that works when certain words are written, like in chatbots or similar things. I mean, the risk that we overlook something or miss something […] is very high.” (TH02). 3.4 AI as a third element (B) : Further, the potential drawback of AI as an additional agent in therapy was discussed, with concerns that it could disrupt the therapeutic relationship by causing confusion or conflicting interpretations, which can complicate the process and detract from the patient-therapist connection: “There is then a third element that may influence the therapeutic relationship. [....] Then we have triangulation, […], which I think is complex. In principle, it's a bit like two therapists working on the same patient.” (TH01). 3.5 (Commercial) conflicts of interest (B) : Another potential disadvantage might arise from commercial interests in adopting AI technologies. Concerns were expressed about profit-driven motives and exploitation of sensitive data for commercial gain, as technologies are rarely developed by non-profit companies. 4. Adopters At the adopter level , nine distinct categories emerged as facilitators or barriers. 4.1 Affinity for technology (F) : Participants highlighted the importance of personal affinity for technology in its adoption, noting that individual user preferences affect how much they want to get involved in technology use and deal with its additional demands. 4.2 Trust and openness (F) : Trust in and openness towards AI is crucial, especially in intimate or emotionally charged contexts, given the sensitive nature of the topics addressed in psychotherapy and the data shared with these technologies. 4.3 Training requirement (F) : Since integrating new technologies can be time-consuming, therapists and patients suggested that training, through videos or practical workshops, could be an essential support mechanism. This was considered especially important in smaller practices or clinics where staff need to be trained quickly to avoid staff shortfall. 4.5 Technology understanding (F) : Closely related to the need for training is the level of users’ knowledge or understanding to use AI technologies effectively. Therapists expressed the need to understand the technology to recommend it confidently to patients or to answer patients’ questions. While patients found it essential that the therapists understand the technology, they did not emphasize the importance of their own understanding. “[…] Therapists […] [should] also [be] familiar with it. So that they can also give me recommendations or perhaps explain how to use it. And that they're not just at the level of saying: 'Well, there's this app, you can use it, it's good, but I don’t really know much else about it'.” (PA06). 4.5 Lack of control (B) : Therapists were concerned about the potential loss of control in the therapeutic process, such as their influence on the patient. Integrating AI into psychotherapy raised questions about maintaining control and ensuring it aligns with the therapist’s approach and methods. 4.6 Lack of human contact (B) : Another emerging concern for both user groups was the lack of human contact, especially the absence of social and emotional connection. The loss of human connection, which is generally perceived as a main mechanism in psychotherapy, may hinder patients’ ability to open up and engage in meaningful and healing therapeutic work. “So, I wouldn't feel seen or acknowledged at all because I would just know: ‘Okay, it's just an algorithm.’ Artificial intelligence or not. What I would miss is the living counterpart, emphasis on living.” (PA03). 4.7 Changes in the importance of conventional psychotherapy (B) : Further, concerns were expressed that integrating AI could diminish the perceived value of and demand for in-person psychotherapy, leading to fewer therapy slots and steering patients away from traditional, personalized care. "I can only put it plainly: there’s more of a concern that they’ll think, ‘Oh, great. A cheap, easy, and quickly available alternative - then we can cut costs elsewhere’." (PA03). 4.8 Technology dependency (B) : Both therapists and patients were concerned about the potential for technology over-reliance. With its constant availability, there are fears that it could lead to an unhealthy dependence on devices rather than fostering real-world coping mechanisms. "What I mentioned earlier as a pro argument could also be seen as a con argument here: the shift from independence from the therapist to dependence on a device." (TH18). 4.9 Persistence of safety-seeking and avoidance behavior (B) : Finally, the use of such technologies might reinforce avoidance or safety behaviors in patients. The ease of using AI, especially for patients with social anxiety or phobias, could encourage avoidance of real-world challenges, hindering therapeutic improvements. While AI may offer immediate relief by eliminating the need for human interaction or enhancing perceived control through safety behaviors such as using conversational agents, it risks distracting from deeper emotional work. “But it can perhaps also lead to patients starting to avoid contact or the relationship difficulties that only become apparent in therapy through shared experience, so that they don't arise in the first place.” (TH02). 5. Organizations At the organizational level, six categories emerged as either facilitator, barrier, or mixed factor. 5.1 Pre-use testing opportunities (F) : Opportunities to interact with and test AI technologies prior to their use are essential. Such opportunities can be created, for instance, through organizational support from the top down. This would allow individuals to test the technology in a neutral setting before using it with patients. 5.2 Other incentives (F) : Both patients and therapists believed that additional incentives should be introduced to support the integration of AI technologies into psychotherapy, such as a dedicated hotline for addressing questions that arise during and after the implementation process. 5.3 Resource and workload considerations (B) : Therapists expressed concerns that AI integration would initially create more effort than relief, with the required resources, particularly the time investment, frequently described as a barrier. Hence, they stressed the need for AI adoption to cause minimal additional workload, considering the already demanding nature of their profession. “For me, […] this familiarization period would be very, very important, because I think it's very important for us, but above all, for the patients, that you don't somehow spend a whole hour of therapy on it [the familiarization period], so to speak, to deal with it first.” (TH17). 5.4 Differences in therapy methods (M) : Therapists usually specialize in a specific therapeutic approach, which may vary in compatibility with AI technologies. AI might be more easily integrated into structured approaches like cognitive behavioral therapy, where predefined methods and tasks are common. However, more dynamic and individual approaches, such as psychodynamic or systemic therapies, may face challenges since their change mechanisms rely on countertransference, childhood experiences, and deeper emotional engagement. “I believe that behavioral therapy has far fewer problems than psychodynamic psychology in the whole complex that is being discussed here. And I think psychoanalysis is probably completely excluded.” (TH06). 5.5 Differences in the type of institution (M) : The type of institution providing therapy also impacts AI adoption, as therapists work in diverse settings from private practices to large clinics. Clinics could be more inclined to adopt AI technologies, as they tend to have the resources and infrastructure to integrate new technologies more easily. However, adoption decisions may also be influenced by financial constraints, administrative structures, and the overall attitude within the institution. “At least in my clinic I can imagine that the clinic management or the head physician could be interested in this, because they tend to have such a market-oriented awareness and say that innovations and things like that are good and can certainly be incorporated.” (TH11). However, the downsides of AI adoption in clinics with top-down decision-making were also discussed: “So, if I work in a clinic or in an outpatient clinic that I don't manage myself, so to speak, then I'm dependent on the decisions of the person in charge. Whereas if I have my own outpatient practice, then I'm more independent and can introduce it more easily.” (TH18). 5.6 Digitalization (M) : Participants felt that digitalization in psychotherapy is not yet advanced enough for the introduction of AI tools. Concerns were raised about the slow digital transformation in healthcare, with many institutions still lacking basic infrastructure like Wi-Fi. “I believe that we haven't yet reached the point in the development of digitalization where it really saves time, where it [technology adoption] becomes automatic and where people have internalized it in such a way that they can say: ‘Okay, it [technology adoption] makes my work easier.” (TH09). 6. Wider system At the wider system level, six categories emerged as either facilitator, barrier, or mixed factor. 6.1 Cost coverage (F) : The prevailing topic was cost coverage. Therapists expressed the need for reimbursement strategies, and both user groups advocated for standardized insurance practices to ensure equitable access for patients. 6.2 Economic feasibility (F) : According to participants, the coverage of development and maintenance costs of AI systems must be ensured to facilitate their adoption. Further, it was noted that economic feasibility should not be pursued at the expense of quality and long-term value. 6.3 Liability (B) : The issue of liability for AI outcomes emerged as another concern, highlighted by a hypothetical scenario. “Let's say I have suicide prevention software, and it somehow works better, ten times better than a therapist […] [in detecting risks]. But still, sometimes people die by suicide. The question then is: Who is somehow liable? Who bears the risk?” (TH08). 6.4 Data protection measures (B) : Participants expressed significant concerns about data security, emphasizing risks related to data breaches, unclear data processing, and storage. 6.5 Regulations and regulatory environment (M) : Regulations to guide the use of AI in psychotherapy were deemed necessary. Participants emphasized the importance of clear legal frameworks, both nationally and internationally, through centralizing regulatory efforts to ensure data protection and quality standards and prevent misuse of sensitive information. However, some voiced concerns about overregulation hindering flexibility. “If this is always linked to any guidelines, then it may be that the whole thing becomes quite a cage as far as the application is concerned.” (TH09). 6.6 Transformations of existing structures (M) : Participants from both groups emphasized the need for a thoughtful approach between using AI to complement existing therapies and questioning whether its use would improve care or just address supply shortages short-term. “But we could perhaps become even faster. We could become even more productive. Perhaps we could see even more patients. Treat even more. [...] Whether that is necessarily a good thing remains to be seen.” (TH10). 7. Embedding and adaptation over time Regarding the embedding and potential adaptations over time, two categories were derived. 7.1 Ongoing scientific validation (F) : Participants highlighted the necessity of ongoing research to validate the efficacy of AI technologies. While initial studies exist, the consensus was that widespread implementation requires constant research, including pilot projects and rigorous testing to meet clinical standards. 7.2 Speed of the implementation process (M) : Opinions were mixed regarding the pace of AI integration into psychotherapy. Some participants acknowledged the unpredictability of technological advancements, suggesting that the adoption process could proceed rapidly. This is particularly true for simpler and more general AI technologies, such as those for text or image processing, which are projected to be implemented within the next three to five years. However, they anticipated that more specialized and niche applications, such as standalone AI-enabled technologies for VR applications in exposure therapy, take longer to implement, approximately ten years. Further, systemic challenges, including legislative and funding uncertainties, were highlighted as significant obstacles to AI implementation. “This will probably take decades until Germany might actually reach the point where they say: ‘OK, AI-supported psychotherapy, […] maybe’. But only if all the framework conditions are 100% perfectly balanced and clarified." (PA05). 8. NASSS extension – Time of use As the aim of this research is to gain a comprehensive understanding of patients’ and therapists’ perspectives on the use of AI in psychotherapy, it is crucial to identify the stages and time points in the therapeutic process where technologies are perceived most beneficial. Therefore, we extend NASSS with “ time of use ” as a new domain. 8.1 Use in diagnostics and first screening (M) : Participants acknowledged AI’s potential to conduct preliminary diagnostic evaluations by analyzing patient-completed questionnaires or assessments before therapist review. While therapists viewed it as useful to streamline workflows, concerns were raised about over-reliance, particularly for complex cases where personal history and symptom nuances may be overlooked. 8.2 Use (during waiting times) before therapy (M) : Participants saw the potential of AI to offer therapeutic resources during waiting periods or appointment delays, enabling early engagement. However, some emphasized that, especially in the beginning, therapeutic relationships should include personal interaction, which AI cannot replicate. 8.3 Use between therapy sessions (F) : Participants judged the use of AI technologies between therapy sessions positively, considering them as a tool to offer continuity and fill gaps in the process, such as providing tailored exercises. 8.4 Use in aftercare (F) : AI technologies could also serve as a valuable resource for post-therapy support, to maintain continuity, offer security, and ease the transition to self-management, particularly for patients feeling unprepared to navigate challenges independently. Participants highlighted its potential for relapse prevention and supporting the reduction of therapeutic contact. Discussion With the rise of AI technologies and growing pressure on therapists and mental health systems, applying the NASSS framework to understand both patients’ and therapists’ perspectives offers valuable insights into the challenges and needs of AI adoption. Although patients and therapists represent distinct user groups with different backgrounds, their discussions revolved around largely overlapping topics, with clear differences in only two areas: only therapists highlighted the need to minimize AI-related workload due to limited time and resources, and only patients emphasized AI’s potential to improve efficiency. Notably, nearly half of all codes were allocated to the technology domain (such as its features and functionality) and the adopter domain (related to users’ role), suggesting these areas are of great interest. The discussion highlights the most frequently cited categories, grouped into four broad areas: First, the findings indicate that heterogeneity in psychotherapy , including the specific type of mental health condition , is an important factor influencing AI adoption. The positive attitudes toward AI technology for depression and anxiety disorders align with trends in research and technological development, which predominantly focus on these conditions (e.g., 8 , 11 , 45 ). The reluctance to consider AI technologies for more complex mental illnesses, such as psychotic disorders, aligns with their broader hesitation toward their use in severe symptomatology . In these cases, such as suicidality, hesitation was only alleviated when the AI was intended solely for supplementary support or early detection rather than for use during acute phases or as a standalone solution. Heterogeneity at the organizational level, including different therapeutic approaches , also needs to be considered. Consistent with our findings, prior research suggests that therapists' theoretical orientations influence their acceptance of new technologies, with psychoanalytic therapists showing lower adoption rates 46 . Similarly, we found that AI integration is more compatible with CBT due to its structured, manualized, and rule-based methodologies. Another key barrier in this context is the variation in organizational and institutional environments therapists work in. Therapists noted that in clinical settings, decisions regarding technology adoption are often outside their control and depend on other stakeholders, such as clinic management. Participants believed that while clinicians and management are more motivated to adopt technologies, administrative stakeholders tend to hinder adoption due to investment costs. In contrast, therapists in private practice have greater autonomy but face the challenge of managing adoption on their own. Second, user-specific human factors also played a pivotal role in AI adoption, with the lack of human contact emerging as an especially critical aspect. This supports previous research emphasizing the need to preserve the human element in patient-therapist interaction when adopting AI technologies, especially in the emotionally sensitive field of mental healthcare 33 , 47 . Such concerns stem from fears that AI could undermine authentic human interaction, potentially disrupting meaningful therapeutic relationships 31 , 48 , 49 . For instance, meaningful relationships are shaped by emotional shifts, like countertransference in psychodynamic psychotherapy. Therapists were hesitant that AI technologies could address these shifts appropriately, stressing the need for human responsiveness. The lack of this human element may hinder therapeutic progress, given the well-established importance of the therapeutic relationship in psychotherapy outcomes 50 . Despite the growing body of research on AI systems establishing a human bond and their promising advancements 51 , such as enhancing social presence in chatbot-based interactions to foster rapport 52 , our findings underscore the continued importance that patients and therapists place on human contact. Furthermore, AI’s role as a supplementary tool emerged as a key facilitator. Participants agreed that human therapists are irreplaceable in the therapeutic process, and that AI technologies, such as AI-enabled mental health apps suggesting interventions, should serve only as a supplement, aligning with existing literature 11 , 53 – 55 . Moreover, customization to users’ needs is one strategy that can help address concerns about the lack of human connection. In our study, customization includes adapting to patients’ therapy experience and knowledge, which helps avoid repetition and tailors advice to prevent suggestions that may worsen the situation. This personalized approach might foster a sense of connection and relevance, thereby mitigating concerns that technology may hinder the human element of therapy. The importance of customization is reflected in our findings, as users viewed customization as a critical feature, in line with previous research 28 . While the literature often focuses on technology customization for patients, for instance, to strengthen the therapeutic relationship by personalizing the therapy style and avatar of chatbots 56 , less attention is given to therapists’ perspectives. Our therapists stressed that customization on their end involves aligning technologies also with their practice, ensuring it is tailored to specific tasks that support their work and maintain authenticity, without compromising their therapeutic framework. Ideally, tailoring the technology to each user's needs enhances its usability across different user groups , a key facilitator highlighted in the literature, particularly regarding age 31 , 57 . Studies have found hesitations towards AI technologies in older adults 58 , while younger patients tend to show greater engagement with technologies 59 . However, our findings even extend this by identifying additional sociodemographic factors, such as the patient’s financial situation, cultural background, and language requirements. Third, the accessibility of AI technologies is also a crucial consideration. Participants emphasized the benefit of continuous support and flexible access to therapy content, particularly across locations and times. Other research also underscores the positive perception of accessible AI technologies, such as those via smartphones, which can reach patients in high-risk situations 54 and offer 24/7 support 60 . However, in our study, participants had a more nuanced view: they noted that constant accessibility could lead to dependency , making the absence of AI technology feel unsettling. This dependency could even exacerbate symptoms by fostering compulsive behaviors, such as feeling compelled to engage excessively with technology. As existing research primarily focuses on the harm of dependency on chatbots 61 , our findings extend this concern to AI applications in psychotherapy in general, emphasizing the broader risk. Additionally, both user groups stressed the risk that AI technologies, through individualized and non-confrontational interactions, may reinforce safety-seeking or avoidance behavior by reducing the exposure to therapeutically relevant challenges. In psychotherapy, such challenges – like deliberate ambiguous or paradoxical therapist responses – are often critical for emotional processing and behavioral change. This concern was especially noted in relation to anxiety disorders, particularly social anxiety. While initially, since it has been shown that individuals with social anxiety tend to disclose more information to conversational agents than to human counterparts 62 , 63 , this might seem beneficial by promoting communication and engagement, it might hinder therapeutic process by limiting therapeutic opportunities to engage in real social interactions 64 . Fourth, different structures surrounding AI adoption also play a pivotal role. Having a robust digital infrastructure is the foundation of AI adoption, however, the current infrastructure in both clinical and private practice settings remains insufficient. While the importance of adequate digital infrastructure in healthcare is widely recognized 65 , in practice, Germany lags behind global leaders in the development of AI technologies for mental health disorders 8 and other European countries in the digitalization of medical services 66 . This lack of favorable conditions compounds another challenge identified by therapists: resource and workload constraints . Adopting new technologies requires time and resources, and both our findings and prior studies indicate that therapists are rather reluctant to engage with time-intensive digital implementations 32 . Dealing with poor digital infrastructure adds workload and complicates the AI adoption process. It also undermines other facilitators, such as pre-use testing opportunities and comprehensive training , that have been linked to facilitating adoption 19 , 67 , 68 . Further, financial and structural transformations at the wider-system level, such as cost coverage , could help overcome technology and adopter-level barriers, for instance by ensuring accessibility for diverse user groups. Cost coverage by health insurance companies is a prerequisite for patients to adopt AI technologies 34 . Additionally, therapists noted the need for affordable technologies and reimbursement 69 . However, participants indicated that meeting cost coverage needs might necessitate a restructuring of current reimbursement mechanisms. Moreover, participants anticipated that system-level structures, such as patient referral pathways, would undergo a transformation , as AI could increase access to therapy, allowing for a larger number of patients to be treated. This, in turn, raised some concerns, especially regarding the treatment quality, with the risk that these changes might offer only a temporary increase in access without addressing the underlying systemic issues, making them a “quick fix” rather than a lasting solution, and psychotherapy becoming less patient-centered and rather driven by economic efficiency. Finally, technology embedding and adaptation over time have been largely overlooked in existing literature 35 , 37 . The majority of our participants expressed uncertainty about the speed of the adoption process and refrained from making generalizations across AI technologies. Instead, they anticipate slower adoption for more specialized and niche applications, while simpler and more general AI technologies are believed to be implemented quicker. Our findings highlight some important implications for the development, design, and implementation of AI technologies and further research. First, addressing the specific technological features valued by patients and therapists can support adoption 70 . Particularly, features enhancing patient autonomy, such as psychoeducational content to aid patients in understanding their conditions, self-help exercises like meditation or guided imagery, and diaries, are deemed most useful. Notably, a common characteristic of these features is their utility when therapists are unavailable and during pre-therapy phases, including diagnostic and initial screening, or while awaiting treatment. Second, aligning the purpose of the AI technology with users’ preferred timing of use is critical to reduce hesitation. Participants in our study identified four key timepoints where they believed that AI could support and enhance the therapeutic process. At each of these stages – diagnostics and screening, before therapy, between therapy sessions, and in aftercare – AI’s use would need to be tailored to the specific needs and goals of that phase to be effective and well-received. Third, there was a substantial overlap in the topics discussed by patients and therapists. While addressing the distinct needs of each user group is essential for AI adoption, targeting shared barriers may simplify adoption and reduce complexity for both. Finally, our findings of mixed categories where the same aspect can be perceived as a barrier and facilitator in almost all NASSS domains highlight the complexity on every level of technology adoption. Understanding these dual-role factors is crucial for developing strategies that maximize their facilitative potential while minimizing drawbacks. For instance, the need for clear legal regulations for technology adoption was emphasized among participants and is also highlighted in the literature 7 , 27 . Regulations should clarify the use and storage of sensitive data, delineate tasks appropriate for AI technologies against those that must be performed by humans, and determine whether the use of AI technology should be mandatory or integrated in treatment guidelines. However, regulation should not limit therapists’ treatment flexibility and hinder implementation. Future research is needed to better understand how regulatory frameworks can support the integration of AI while preserving ethical standards and the clinician’s role in therapy. Balancing these concerns is therefore key for fostering both ethical safeguards and innovation. This study has several limitations: First, the sample was imbalanced, with three focus groups with therapists and only two with patients. This occurred because participant recruitment followed the principle of data saturation 71 , meaning that data collection continued until no new topics emerged. The therapist focus groups were conducted first due to earlier sign-ups resulting in fewer patient focus groups, as these started later in the process, and saturation was achieved after five focus groups. Second, our relatively young study sample may reflect a more technology-engaged group 31 , 59 , potentially skewing perceptions. Third, our study only included participants from across Germany; hence, their experiences and perceptions regarding the digital infrastructure and healthcare systems may not be generalizable to other countries. Healthcare policies and technological advancements differ widely across regions, with countries like the U.S. and China leading in AI development for mental healthcare 8 . These regional variances in healthcare systems and regulations may influence how AI technologies are perceived and adopted into mental health treatment, limiting the generalizability of our findings to other contexts with different healthcare frameworks. Conclusion Overall, the study highlights that AI adoption in mental healthcare is a complex and multifaceted issue, influenced by the subjectivity and heterogeneity of mental illnesses, and the diverse therapeutic approaches and workplace environments. Participants’ needs, viewed as facilitating factors, emerged in every domain of the technology adoption process, and when appropriately leveraged, can ease AI adoption. However, given the general low use intentions of AI in mental healthcare, it is essential to design AI solutions that not only align with user needs but also alleviate their concerns. Addressing these barriers proactively holds the potential to overcome hesitation, realize the full potential of AI technology, and consequently, improve mental healthcare by mitigating the treatment gap. Declarations Competing interests: All authors declare no financial or non-financial competing interests. Funding: This study was funded by a grant from the Volkswagen Foundation (Grant #: 98525). The funder played no role in study design, data collection, analysis, and interpretation of data, or the writing of this manuscript. Author Contribution J.C., S.G., and A.-K.K. conceptualized and design the study. J.C. and D.E. recruited participants and collected the data in the focus groups. J.C., D.E. and I.S. analyzed the focus group transcripts. J.C. wrote the original draft of the manuscript. E.L., D.E., I.S., S.G., and A.-K. K. contributed to the writing, interpretation, and editing of the manuscript. All authors read and approved the final manuscript. Data Availability The discussion guide, COREQ checklist and the description of AI and AI-enabled technologies in psychotherapy given to participants prior to the study is available on OSF (https://osf.io/9rt52/?view_only=a1a4b6a50d8f4d62b3b685eee6651a9b). The codebook and exemplary codes supplementing the study results are available in the online Supplementary Information. References Qin, X. & Hsieh, C.-R. Understanding and Addressing the Treatment Gap in Mental Healthcare: Economic Perspectives and Evidence From China. Inquiry. 57, 0046958020950566 (2020). Roberts, T. et al. Reconceptualising the treatment gap for common mental disorders: a fork in the road for global mental health? Br. J. Psychiatry 221, 553–557 (2022). World Health Organization. World Mental Health Report: Transforming mental health for all . https://www.who.int/publications/i/item/9789240049338 (2022). Deutsches Ärzteblatt. Psychotherapie: Längere Wartezeiten auf Therapieplatz seit Strukturreform [Psychotherapy: Longer waiting times for therapy places since structural reform] . https://www.aerzteblatt.de/news/psychotherapie-laengere-wartezeiten-auf-therapieplatz-seit-strukturreform-cbef13cc-5ae 1-4ae6-bf04-58bf00edee58 (2023). Subotic-Kerry, M. et al. While they wait: A cross-sectional survey on wait times for mental health treatment for anxiety and depression for Australian adolescents. BMJ open. 15, e087342 (2024). Wainberg, M. L. et al. Challenges and Opportunities in Global Mental Health: a Research-to-Practice Perspective. Curr. Psychiatry Rep. 19, 28 (2017). Kleine, A.-K. et al. AI-enabled clinical decision support tools for mental healthcare: A product review. Artif. Intell. Med. 160, 103052 (2025). Kleine, A.-K., Lermer, E., Cecil, J., Heinrich, A. & Gaube, S. Advancing mental health care with AI-enabled precision psychiatry tools: A patent review. Comput. Hum. Behav. Rep. 12, 100322 (2023). Olawade, D. B. et al. Enhancing mental health with Artificial Intelligence: Current trends and future prospects. J. Med. Surg. Public Health 3, 100099 (2024). Cecil, J., Kleine, A.-K., Lermer, E. & Gaube, S. Mental health practitioners’ perceptions and adoption intentions of AI-enabled technologies: an international mixed-methods study. BMC Health Serv. Res. 25, 556 (2025). Beg, M. J., Verma, M., M., V. C. K. M. & Verma, M. K. Artificial Intelligence for Psychotherapy: A Review of the Current State and Future Directions. Indian J. Psychol. Med. (2024) doi: 10.1177/02537176241260819 . Stade, E. C. et al. Large language models could change the future of behavioral healthcare: a proposal for responsible development and evaluation. Npj Ment. Health Res. 3, 12 (2024). Squires, M. et al. Deep learning and machine learning in psychiatry: a survey of current progress in depression detection, diagnosis and treatment. Brain Inform. 10, 10 (2023). Rollwage, M. et al. Using Conversational AI to Facilitate Mental Health Assessments and Improve Clinical Efficiency Within Psychotherapy Services: Real-World Observational Study. JMIR AI 2, e44358 (2023). Koh, J., Tng, G. Y. Q. & Hartanto, A. Potential and Pitfalls of Mobile Mental Health Apps in Traditional Treatment: An Umbrella Review. J. Pers. Med. 12, 1376 (2022). Kleine, A.-K., Kokje, E., Lermer, E. & Gaube, S. Attitudes Toward the Adoption of 2 Artificial Intelligence–Enabled Mental Health Tools Among Prospective Psychotherapists: Cross-sectional Study. JMIR Hum. Factors 10, e46859 (2023). Gutierrez, G., Stephenson, C., Eadie, J., Asadpour, K. & Alavi, N. Examining the role of AI technology in online mental healthcare: opportunities, challenges, and implications, a mixed-methods review. Front. Psychiatry 15, 1356773 (2024). Beilharz, F., Sukunesan, S., Rossell, S. L., Kulkarni, J. & Sharp, G. Development of a Positive Body Image Chatbot (KIT) With Young People and Parents/Carers: Qualitative Focus Group Study. J. Med. Internet Res. 23, e27807 (2021). Braun, P. et al. Investigating information needs and preferences regarding digital mental health services among medical and psychology students in Germany: A qualitative study. Digit. Health 9, 20552076231173568 (2023). Creed, T. A. et al. Knowledge and Attitudes Toward an Artificial Intelligence-Based Fidelity Measurement in Community Cognitive Behavioral Therapy Supervision. Adm. Policy Ment. Health Ment. Health Serv. Res. 49, 343–356 (2022). Goodspeed, A. et al. Leveraging the utility of pharmacogenomics in psychiatry through clinical decision support: a focus group study. Ann. Gen. Psychiatry 18, 13 (2019). Henshall, C. et al. Implementing a digital clinical decision support tool for side effects of antipsychotics: a focus group study. BMJ Ment Health 22, 56–60 (2019). Radovic, A. et al. Perceived Facilitators of and Barriers to Implementation of a Decision Support Tool for Adolescent Depression and Suicidality Screening: Focus Group and Interview Study. JMIR Ment. Health 8, e26035 (2021). Greenhalgh, T. et al. The NASSS-CAT Tools for Understanding, Guiding, Monitoring, and Researching Technology Implementation Projects in Health and Social Care: Protocol for an Evaluation Study in Real-World Settings. JMIR Res. Protoc. 9, e16861 (2020). James, H. M., Papoutsi, C., Wherton, J., Greenhalgh, T. & Shaw, S. E. Spread, Scale-up, and Sustainability of Video Consulting in Health Care: Systematic Review and Synthesis Guided by the NASSS Framework. J. Med. Internet Res. 23, e23775 (2021). Brown, P. et al. Automated Virtual Reality Cognitive Therapy (gameChange) in Inpatient Psychiatric Wards: Qualitative Study of Staff and Patient Views Using an Implementation Framework. JMIR Form. Res. 6, e34225 (2022). Berardi, C. et al. Barriers and facilitators to the implementation of digital technologies in mental health systems: a qualitative systematic review to inform a policy framework. BMC Health Serv. Res. 24, 243 (2024). Alqahtani, F., Winn, A. & Orji, R. Co-Designing a Mobile App to Improve Mental Health and Well-Being: Focus Group Study. JMIR Form. Res. 5, e18172 (2021). Terra, M., Baklola, M., Ali, S. & El-Bastawisy, K. Opportunities, applications, challenges and ethical implications of artificial intelligence in psychiatry: a narrative review. Egypt. J. Neurol. Psychiatry Neurosurg. 59, 80 (2023). Freitag, B., Uncovska, M., Meister, S., Prinz, C. & Fehring, L. Cost-effectiveness analysis of mHealth applications for depression in Germany using a Markov cohort simulation. Npj Digit. Med. 7, 1–13 (2024). Aafjes-van Doorn, K. Feasibility of artificial intelligence-based measurement in psychotherapy practice: Patients’ and clinicians’ perspectives. Couns. Psychother. Res. 25, e12800 (2024). Mendes-Santos, C., Nunes, F., Weiderpass, E., Santana, R. & Andersson, G. Understanding Mental Health Professionals’ Perspectives and Practices Regarding the Implementation of Digital Mental Health: Qualitative Study. JMIR Form. Res. 6, e32558 (2022). Hogg, H. D. J. et al. Stakeholder Perspectives of Clinical Artificial Intelligence Implementation: Systematic Review of Qualitative Evidence. J. Med. Internet Res. 25, e39742 (2023). Richardson, J. P. et al. Patient apprehensions about the use of artificial intelligence in healthcare. Npj Digit. Med. 4, 1–6 (2021). Abell, B. et al. Identifying barriers and facilitators to successful implementation of computerized clinical decision support systems in hospitals: a NASSS framework-informed scoping review. Implement. Sci. 18, 32 (2023). Ahmed, M. I. et al. A Systematic Review of the Barriers to the Implementation of Artificial Intelligence in Healthcare. Cureus 15, e46454 (2023). Bastoni, S. et al. Factors Influencing Implementation of eHealth Technologies to Support Informal Dementia Care: Umbrella Review. JMIR Aging 4, e30841 (2021). Hummelsberger, P. et al. Insights on the Current State and Future Outlook of AI in Health Care: Expert Interview Study. JMIR AI 2, e47353 (2023). Lim, W. M. What Is Qualitative Research? An Overview and Guidelines. Australas. Mark. J. (2024) doi: 10.1177/14413582241264619 . Tong, A., Sainsbury, P. & Craig, J. Consolidated criteria for reporting qualitative research (COREQ): a 32-item checklist for interviews and focus groups. Int. J. Qual. Health Care 19, 349–357 (2007). GBD 2019 Mental Disorders Collaborators. Global, regional, and national burden of 12 mental disorders in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet Psychiatry 9, 137–150 (2022). Freeman, T. ‘Best practice’ in focus group research: making sense of different views. J. Adv. Nurs. 56, 491–497 (2006). Fereday, J. & Muir-Cochrane, E. Demonstrating Rigor Using Thematic Analysis: A Hybrid Approach of Inductive and Deductive Coding and Theme Development. Int. J. Qual. Methods 5, 80–92 (2006). McHugh, M. L. Interrater reliability: the kappa statistic. Biochem. Medica 276–282 (2012) doi: 10.11613/BM.2012.031 . Herbener, A. B., Klincewicz, M. & Damholdt, M. F. A narrative review of the active ingredients in psychotherapy delivered by conversational agents. Comput. Hum. Behav. Rep. 14, 100401 (2024). Bourla, A. et al. Psychiatrists’ Attitudes Toward Disruptive New Technologies: Mixed-Methods Study. JMIR Ment. Health 5, e10240 (2018). Blease, C. et al. Artificial Intelligence and the Future of Primary Care: Exploratory Qualitative Study of UK General Practitioners’ Views. J. Med. Internet Res. 21, e12802 (2019). Haber, Y., Levkovich, I., Hadar-Shoval, D. & Elyoseph, Z. The Artificial Third: A Broad View of the Effects of Introducing Generative Artificial Intelligence on Psychotherapy. JMIR Ment. Health 11, e54781–e54781 (2024). Ruse, J. N., Schraube, E. & Rhodes, P. Left to their own devices: the significance of mental health apps on the construction of therapy and care. Subjectivity 31, 410–428 (2025). Lambert, M. J. & Barley, D. E. Research summary on the therapeutic relationship and psychotherapy outcome. Psychother. Theory Res. Pract. Train. 38, 357–361 (2001). Morrow, E. et al. Artificial intelligence technologies and compassion in healthcare: A systematic scoping review. Front. Psychol. 13, 971044 (2023). Lee, J., Lee, D. & Lee, J. Influence of Rapport and Social Presence with an AI Psychotherapy Chatbot on Users’ Self-Disclosure. Int. J. Human–Computer Interact. 40, 1620–1631 (2024). Babu, A. & Joseph, A. P. Artificial intelligence in mental healthcare: transformative potential vs. the necessity of human interaction. Front. Psychol. 15, 1378904 (2024). Götzl, C. et al. Artificial intelligence-informed mobile mental health apps for young people: a mixed-methods approach on users’ and stakeholders’ perspectives. Child Adolesc. Psychiatry Ment. Health 16, 86 (2022). Rebelo, A. D., Verboom, D. E., Dos Santos, N. R. & De Graaf, J. W. The impact of artificial intelligence on the tasks of mental healthcare workers: A scoping review. Comput. Hum. Behav. Artif. Hum. 1, 100008 (2023). Vossen, W., Szymanski, M. & Verbert, K. The effect of personalizing a psychotherapy conversational agent on therapeutic bond and usage intentions. in Proceedings of the 29th International Conference on Intelligent User Interfaces 761–771 (Association for Computing Machinery, New York, NY, USA, 2024). doi: 10.1145/3640543.3645195 . Bucci, S. et al. “They Are Not Hard-to-Reach Clients. We Have Just Got Hard-to-Reach Services.” Staff Views of Digital Health Tools in Specialist Mental Health Services. Front. Psychiatry 10, (2019). Fritsch, S. J. et al. Attitudes and perception of artificial intelligence in healthcare: A cross-sectional survey among patients. Digit. Health 8, 20552076221116772 (2022). Connolly, S. L. et al. Patient and provider predictors of telemental health use prior to and during the COVID-19 pandemic within the Department of Veterans Affairs. Am. Psychol. 77, 249–261 (2022). Aktan, M. E., Turhan, Z. & Dolu, İ. Attitudes and perspectives towards the preferences for artificial intelligence in psychotherapy. Comput. Hum. Behav. 133, 107273 (2022). Laestadius, L., Bishop, A., Gonzalez, M., Illenčík, D. & Campos-Castillo, C. Too human and not human enough: A grounded theory analysis of mental health harms from emotional dependence on the social chatbot Replika. New Media Soc. 26, 5923–5941 (2024). Hu, B. How social anxiety leads to problematic use of conversational AI: The roles of loneliness, rumination, and mind perception. Comput. Hum. Behav. 145, 107760 (2023). Kang, S. & Gratch, J. Virtual humans elicit socially anxious interactants’ verbal self-disclosure. Comput. Animat. Virtual Worlds 21, 473–482 (2010). Rodebaugh, T. L., Holaway, R. M. & Heimberg, R. G. The treatment of social anxiety disorder. Clin. Psychol. Rev. 24, 883–908 (2004). Hassan, M., Kushniruk, A. & Borycki, E. Barriers to and Facilitators of Artificial Intelligence Adoption in Health Care: Scoping Review. JMIR Hum. Factors 11, e48633 (2024). Majcherek, D., Hegerty, S. W., Kowalski, A. M., Lewandowska, M. S. & Dikova, D. Opportunities for healthcare digitalization in Europe: Comparative analysis of inequalities in access to medical services. Health Policy 139, 104950 (2024). Golden, G. et al. Applying artificial intelligence to clinical decision support in mental health: What have we learned? Health Policy Technol. 13, 100844 (2024). Lambert, S. I. et al. An integrative review on the acceptance of artificial intelligence among healthcare professionals in hospitals. Npj Digit. Med. 6, 1–14 (2023). Lee, E. E. et al. Artificial Intelligence for Mental Health Care: Clinical Applications, Barriers, Facilitators, and Artificial Wisdom. Biol. Psychiatry Cogn. Neurosci. Neuroimaging 6, 856–864 (2021). Schueller, S. M., Neary, M., O’Loughlin, K. & Adkins, E. C. Discovery of and Interest in Health Apps Among Those With Mental Health Needs: Survey and Focus Group Study. J. Med. Internet Res. 20, e10141 (2018). Saunders, B. et al. Saturation in qualitative research: exploring its conceptualization and operationalization. Qual. Quant. 52, 1893–1907 (2018). Additional Declarations No competing interests reported. Supplementary Files SupplementaryInformation.pdf Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 07 Dec, 2025 Reviews received at journal 05 Dec, 2025 Reviewers agreed at journal 14 Nov, 2025 Reviewers agreed at journal 13 Nov, 2025 Reviewers agreed at journal 10 Nov, 2025 Reviewers agreed at journal 29 Sep, 2025 Reviewers agreed at journal 03 Aug, 2025 Reviewers agreed at journal 02 Aug, 2025 Reviews received at journal 22 Jun, 2025 Reviewers agreed at journal 01 Jun, 2025 Reviewers invited by journal 01 Jun, 2025 Editor assigned by journal 12 May, 2025 Submission checks completed at journal 29 Apr, 2025 First submitted to journal 28 Apr, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6546781","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":449830136,"identity":"c3dba56e-4227-45aa-9f03-b5ca4d0a85da","order_by":0,"name":"Julia Cecil","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA9ElEQVRIie3RsQrCMBAG4AuBTtFuIlT0FSKOgs9iKejiUHARFI0U4iK4VvAh9A1SAk6+gYvuXcTFoYhJpeBidBTMP+TuIB8XCICNzQ8GMcxE3mE8P+la1gcxElQQFFFdnU9Eo6Lh1a8Ijv25uEGn7i4QH4eZBKdyoJCODDtinyVLCFpVifhxzRXxBhRtDmYiCAh/ixUpsSfBJf5hSwZipsmQZAW5m4lUW7pUEUycgjADWZ6ZrNGgGUsUeWveJ47XC5PN/j1pLgJ5SUedhruSyTXM2nXXC3andGIg+Qvoc8AAUf4j4j0AaLwOikxNl21sbGz+NA9dHEyTbiKHnAAAAABJRU5ErkJggg==","orcid":"","institution":"LMU Munich","correspondingAuthor":true,"prefix":"","firstName":"Julia","middleName":"","lastName":"Cecil","suffix":""},{"id":449830137,"identity":"4c6bd868-4a2c-4686-8828-1d29c1aeda5f","order_by":1,"name":"Insa Schaffernak","email":"","orcid":"","institution":"Technical University of Applied Sciences Augsburg","correspondingAuthor":false,"prefix":"","firstName":"Insa","middleName":"","lastName":"Schaffernak","suffix":""},{"id":449830138,"identity":"39b71ddb-a720-4009-a9ea-7ab2da29175f","order_by":2,"name":"Danae Evangelou","email":"","orcid":"","institution":"LMU Munich","correspondingAuthor":false,"prefix":"","firstName":"Danae","middleName":"","lastName":"Evangelou","suffix":""},{"id":449830139,"identity":"e3fba172-107b-4471-98f0-d24244932dfe","order_by":3,"name":"Eva Lermer","email":"","orcid":"","institution":"LMU Munich","correspondingAuthor":false,"prefix":"","firstName":"Eva","middleName":"","lastName":"Lermer","suffix":""},{"id":449830140,"identity":"24cd28a8-fb53-4ca2-a9f8-42ccff723716","order_by":4,"name":"Susanne Gaube","email":"","orcid":"","institution":"University College London","correspondingAuthor":false,"prefix":"","firstName":"Susanne","middleName":"","lastName":"Gaube","suffix":""},{"id":449830141,"identity":"bcfc66b2-8d44-4725-af24-5c9cb8b0cdf8","order_by":5,"name":"Anne-Kathrin Kleine","email":"","orcid":"","institution":"LMU Munich","correspondingAuthor":false,"prefix":"","firstName":"Anne-Kathrin","middleName":"","lastName":"Kleine","suffix":""}],"badges":[],"createdAt":"2025-04-28 10:23:25","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6546781/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6546781/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":81962360,"identity":"dd7521cd-aa1c-456c-a82d-9e7bc6adb12f","added_by":"auto","created_at":"2025-05-05 11:11:12","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":52678,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eThe NASSS framework specified for AI in psychotherapeutic treatment, adapted from Greenhalgh et al. \u003c/em\u003e\u003csup\u003e24\u003c/sup\u003e\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6546781/v1/b2856b8d1c31a363803b2a45.png"},{"id":81962362,"identity":"97ca8125-c9a3-4c8b-8f8f-5b3d248991e3","added_by":"auto","created_at":"2025-05-05 11:11:12","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":116992,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eTree map with categories regarding AI adoption in psychotherapy clustered along the NASSS domains\u003c/em\u003e.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6546781/v1/ac5644105cd254adf4e050a3.png"},{"id":81966166,"identity":"8ce34448-fb7c-4930-90f4-d8d9e56bead1","added_by":"auto","created_at":"2025-05-05 11:35:12","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1706457,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6546781/v1/545f087b-d317-46fa-8f1f-ff86c1437fad.pdf"},{"id":81962363,"identity":"f76d7009-3e31-45d3-bb4a-4fe2d3a2d727","added_by":"auto","created_at":"2025-05-05 11:11:12","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":316836,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryInformation.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6546781/v1/222eeac69795e4eba220fc30.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Navigating the complexity of AI adoption in psychotherapy by identifying key facilitators and barriers","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe treatment gap in mental healthcare, resulting from a discrepancy between the need for treatment and its availability, remains a significant global challenge\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. One in eight people is likely to develop a mental disorder during their lifetime\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e. However, with a global median of only 13 mental health professionals per 100,000 people, each would theoretically be responsible for over 7,000 individuals, including approximately 960 requiring care\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e. This shortage is reflected in long waiting times, for example, in Germany, patients wait an average of 14.5 weeks for an initial appointment\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e, with other countries showing similar patterns\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. This unrealistic demand for and shortage of mental health professionals, often coupled with inadequate coordination among psychotherapy providers, hinders optimal care\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e,\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e. Several additional factors contribute to inadequate treatment, including access barriers (e.g., travel costs), fear of stigmatization, and the misinterpretation of mental health symptoms as normal responses to social and economic challenges, rather than as treatable conditions\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe advent of digital health technologies enabled by artificial intelligence (AI) presents new opportunities to help bridge the treatment gap and transform mental healthcare. AI-enabled technologies include tools for screening, diagnosing, predicting, and treating mental health conditions, as well as solutions designed to enhance therapists' performance and streamline administrative tasks\u003csup\u003e\u003cspan additionalcitationids=\"CR8 CR9\" citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e. While technologies for screening, diagnosing, feedback and practice management primarily focus on decision support and efficiency improvements, AI-enabled treatment applications actively engage both patients and therapists in the psychotherapeutic process. This paper focuses on the latter category of applications within the psychotherapeutic treatment process, emphasizing their dynamic interaction with both user groups \u0026ndash; a feature that distinguishes them from other application areas. These technologies leverage advanced algorithms to customize treatment\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e. For instance, they can support therapists\u0026rsquo; work by providing psychoeducation or therapeutic interventions specifically tailored to individual patient\u0026rsquo;s needs\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. Oftentimes, these AI-enabled treatment tools focus on depression and anxiety\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e,\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e. For instance, Natural Language Processing (NLP) models can assist in selecting the most effective therapeutic approach for treating depression, such as cognitive behavioral therapy or drug treatment, tailored to the individual patient\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e. Overall, AI-enabled technologies have the potential to enhance the therapists\u0026rsquo; work experience by reducing assessment times\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e, while providing improved and more timely support for patients\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e,\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\n\u003ch3\u003eAI adoption in psychotherapy\u003c/h3\u003e\n\u003cp\u003eDespite these possibilities, both patients and therapists demonstrate reservations about these technologies. Mental health professionals show a low intention to use AI-enabled technologies, especially patient-centered tools, including those used for treatment \u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e,\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e. Among patients, research has demonstrated that dropout rates for AI-enabled psychoeducation applications, for example, range from 5\u0026ndash;34%\u003csup\u003e17\u003c/sup\u003e. This reluctance might highlight a critical issue: much of the existing research on technology adoption in psychotherapy focuses on the design and functionality of AI applications rather than addressing user needs\u003csup\u003e\u003cspan additionalcitationids=\"CR19 CR20 CR21 CR22\" citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e. Understanding user needs and perceptions towards AI technologies is crucial to identify, address, and ideally prevent barriers that could hinder appropriate technology adoption.\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eTheoretical framework for understanding barriers to AI adoption\u003c/h2\u003e \u003cp\u003eRecognizing that many healthcare innovations fail due to a variety of factors, the Nonadoption, Abandonment, and Challenges to the Scale-up, Spread, and Sustainability (NASSS) framework was developed by Greenhalgh et al.\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e. It provides a structural approach to analyze multiple dimensions and complex dynamics shaping the use of AI in healthcare by capturing barriers and facilitators to technology adoption in seven domains (see Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). In mental healthcare, the NASSS framework has been used to guide adoption research on video therapy consultations\u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e, and internet-delivered cognitive behavioral therapy (CBT), and CBT in virtual reality\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe first domain, \u003cem\u003econdition(s) or illness\u003c/em\u003e, focuses on the nature and complexity of health conditions targeted by the technology in question\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e. It further encompasses clinical and sociocultural aspects, such as comorbidities\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e. Research on the adoption of AI technologies designed for mental disorders, such as depression or eating disorders, has mostly focused on already developed specific applications\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e,\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e. However, little research takes a bottom-up approach by first exploring which disorders users consider suitable or not for AI treatment tools.\u003c/p\u003e \u003cp\u003eThe second domain, \u003cem\u003etechnology or technologies\u003c/em\u003e, refers to the technology\u0026rsquo;s characteristics, including their features, functionality, and usability\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e. An established requirement for AI adoption in mental healthcare is the technology\u0026rsquo;s foundation in robust scientific evidence and its capacity for personalization to meet patient's specific needs\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e,\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e. Interestingly, specific features vary in relevance. For instance, tracking and feedback are generally viewed positively, while features like goal setting and relaxation audios are less preferred and may pose barriers to adoption\u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eIn the next domain, \u003cem\u003evalue proposition\u003c/em\u003e, concerns and benefits, such as the perceived desirability, efficiency, and cost-effectiveness of the technology, are examined\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e. From an economic perspective, some applications, such as AI-enabled computer assisted therapy, have been shown to be cost-efficient\u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e. However, the use of digital technologies for the diagnosis and treatment of depression, does not seem to lead to healthcare cost savings\u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe fourth domain, \u003cem\u003eadopter system\u003c/em\u003e, includes users\u0026rsquo; perceptions of adoption and workload considerations, addressing factors such as training needs and individual-level barriers to adoption. On the therapist side, these issues include insufficient training, both before and during the adoption of AI technologies\u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e,\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e,\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e. Additionally, there are concerns about the dehumanization of care and perceived power imbalances between therapists and patients due to shifting dynamics in the triadic relationship of therapists, patients, and technology\u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e. From the patient's perspective, key barriers also include a perceived loss of human connection, and lack of professional oversight\u003csup\u003e\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e,\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eNext, the domain \u003cem\u003eorganization(s)\u003c/em\u003e, addresses the readiness and capacity of an organization for a specific technology\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e. Although limited research addresses barriers at this level in mental healthcare, financial challenges hindering AI adoption were identified in other healthcare contexts\u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e,\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u003c/sup\u003e. Specifically, a lack of funding, costs associated with expert input for labeling and curating AI training data, and technology maintenance create significant barriers to adoption at the organizational level\u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e,\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe sixth domain, \u003cem\u003ewider context\u003c/em\u003e, considers the structure, dynamics, and capacity of the healthcare system in which the innovation is introduced\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e. The domain examines how the new technology interacts with existing practices, policies, and resources. Prior research on AI in psychotherapy identified obstacles such as cost coverage\u003csup\u003e\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e and regulatory issues, specifically the absence of structured implementation frameworks\u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e,\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e,\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe final domain, e\u003cem\u003embedding and adaptation over time\u003c/em\u003e, addresses the challenges and strategies associated with expanding the reach and impact of the innovation beyond initial pilot or research settings\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e. This domain has received limited attention in research\u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e,\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e; however, some factors such as ongoing technology maintenance seem to play a role\u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e,\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eDespite a growing body of research on AI applications in mental healthcare, a gap remains in understanding the adoption barriers from a bottom-up perspective, particularly given the limited findings in this specific field of healthcare. This study, therefore, investigates the needs and barriers from both patient and therapist perspectives through the NASSS framework, addressing areas typically explored through a top-down lens, such as organizational and system-level structures. In doing so, we aim to generate insights that can inform best practices for stakeholders, including developers, helping to proactively address user reservations prior to the development of AI technologies.\u003c/p\u003e \u003c/div\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\n \u003ch2\u003eStudy design\u003c/h2\u003e\n \u003cp\u003eThis qualitative study used focus groups to explore the needs of both patients and therapists for AI-enabled technologies in psychotherapy. Focus groups provide an opportunity for an exploratively in-depth analysis of an area with limited prior research and enable nuanced, contextual understanding of participants\u0026rsquo; perspectives and underlying reasoning\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e39\u003c/span\u003e\u003c/sup\u003e. This pre-registered study (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://osf.io/nzb2k/?view_only=c9fcc933d2e8421c981d73cd740f95fd\u003c/span\u003e\u003c/span\u003e) received ethical approval from the [blinded for review], identifier: 2023-11-V-144-R. It adheres to the 32-item Consolidated Criteria for Reporting Qualitative Research guidelines\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e, see supplementary material on OSF.\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003eRecruitment and Participants\u003c/h3\u003e\n\u003cp\u003eThe study included two participant groups: therapists and patients. Therapists were recruited via emails distributed among psychotherapy training institutions and patients were recruited via flyers in and distribution channels of two German universities and two German patient organizations. Therapists were required to be licensed or undergoing psychotherapy training. Patients were required to be enrolled in psychotherapy and to be diagnosed with either depression or anxiety disorder. These conditions were chosen due to their high global prevalence and burden\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eA total of 32 participants were recruited, comprising 19 therapists and 13 patients. Participant characteristics can be found in Table\u0026nbsp;1. Each focus group consisted of four to six participants, with therapists and patients placed in separate groups. Using homogenous groups enhances the external validity of the research as it allows for in-depth discussions between participants, providing each group with space tailored to their specific experiences and perspectives\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e. Additionally, this approach helps mitigate any potential power imbalances between therapists and patients, creating a more open and balanced discussion environment\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n \u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;Table 1 Sample demographics\u003c/div\u003e\n \u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n \u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n \u003ctable id=\"Taba\" border=\"1\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePatients (n\u0026thinsp;=\u0026thinsp;13)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTherapists (n\u0026thinsp;=\u0026thinsp;19)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eN (%) or M (SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eN (%) or M (SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAge\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e30.23 (10.05)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e33.32 (10.55)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGender\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e10 (76.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e15 (79.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2 (15.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4 (21.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNon-binary/ third gender\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1 (7.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMental disorder\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDepression\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e10 (55.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAnxiety\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5 (27.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOther\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3 (16.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eProfession\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTherapist in training\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e17 (89.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTherapist\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2 (10.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTherapeutic approach\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e(Cognitive) behavioral\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e12 (57.14)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePsychodynamic\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6 (28.57)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePsychoanalytic\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2 (9.52)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSystemic\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1 (4.76)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eWorkplace\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePractice\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3 (10.34)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePrivate practice\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1 (3.45)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSpecialist hospital for psychiatry, psychotherapy, psychosomatic medicine or neurology\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7 (24.18)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRehabilitation clinic\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4 (13.79)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e(University) outpatient clinic\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e10 (34.48)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCommunity mental health center/ counseling center\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1 (3.45)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOther\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3 (10.34)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eProfessional experience (in years)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3.89 (4.78)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"3\" align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eNote\u003c/em\u003e. Combined age for therapists and patients: \u003cem\u003eM\u003c/em\u003e\u003csub\u003eall\u003c/sub\u003e = 32.06, \u003cem\u003eSD\u003c/em\u003e\u003csub\u003eall\u003c/sub\u003e = 10.3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003ch3\u003eData collection\u003c/h3\u003e\n\u003cp\u003eData was collected between March and May 2024. The online focus groups were conducted in German and via Zoom (Zoom Video Communications, Inc.), and lasted 58.36 minutes on average (\u003cem\u003eSD\u003c/em\u003e\u0026thinsp;=\u0026thinsp;2.05, \u003cem\u003emin\u003c/em\u003e\u0026thinsp;=\u0026thinsp;56.23, \u003cem\u003emax\u003c/em\u003e\u0026thinsp;=\u0026thinsp;61.49). Participants signed a data protection and consent form prior to their participation. Additionally, they received a general description of AI and AI-enabled technologies in psychotherapy to facilitate their familiarity with the subject in advance. The description was intended to offer a general overview of the various tasks AI can perform in psychotherapy, rather than concentrating solely on a specific tool. The description can be found in the supplementary material on OSF. At the beginning of each focus group, participants were informed about the purpose and procedure of the discussion and that their participation was voluntary and could be withdrawn at any time. The focus groups proceeded with a brief recap of AI-enabled technologies in mental healthcare to reinforce participants\u0026rsquo; understanding. The focus groups were facilitated by two moderators. One moderator guided the discussions, while the other posted the discussion guide questions in the chat and addressed participants\u0026rsquo; questions. For the questions, a semi-structured discussion guide was used. The discussion guide questions are based on the NASSS framework \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e and related studies in this field (for instance, see \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e). Questions and prompts only differed in the wording to address each groups\u0026rsquo; distinct characteristics, see online Supplementary Information.\u003c/p\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n \u003ch2\u003eData Analysis\u003c/h2\u003e\n \u003cp\u003eThe sessions were recorded and transcribed verbatim using Trint (Trint Limited). Content analysis was conducted with the transcripts using MAXQDA (VERBI Software). Data analysis followed the deductive and inductive approach outlined by Fereday \u0026amp; Muir-Cochrane\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e. First, a code template was deductively developed based on the theoretical structure of the NASSS framework, which provided seven broad coding domains\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e. The template contains the names of the domains with a definition and how to apply them to the transcripts. Second, to ensure the reliability of the codes, two researchers (JC and DE) familiarized themselves with the transcripts and third, the transcripts were coded independently according to the respective seven domains of the NASSS. Overall, complete sentences were used as the smallest coding unit and multiple codes were applied to the same sections if necessary. Subsequently, an inductive approach was employed to identify emerging themes from the data, which were organized within these predefined NASSS domains. In this fourth step, two researchers coded the transcripts independently again. It is worth noting that, during this step, the original NASSS subcategories were coded when explicitly mentioned in the focus groups. However, participants predominantly generated new categories that exceeded existing NASSS subcategories. As a result, we opted not to constrain new categories to the predefined NASSS subcategories, ensuring that the analysis remained true to the participants\u0026rsquo; contributions. Fifth, as new themes emerged, they were grouped together and assigned concise labels. Lastly, after the independent coding, both researchers compared their categories and the names of the categories were refined and merged, resulting in a final codebook. Both researchers then independently coded all transcripts again. The intercoder agreement after this step, measured as a Cohen \u0026kappa; value of 0.73 for segments with at least 90% overlap, indicates a moderate to substantial level of agreement\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e. Any discrepancies in coding during this step were discussed with a third researcher (IS) and resolved through consensus.\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003eData availability\u003c/h3\u003e\n\u003cp\u003eThe discussion guide, COREQ checklist and the description of AI and AI-enabled technologies in psychotherapy given to participants prior to the study is available on OSF (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://osf.io/9rt52/?view_only=a1a4b6a50d8f4d62b3b685eee6651a9b\u003c/span\u003e\u003c/span\u003e). The codebook and exemplary codes supplementing the study results are available in the online Supplementary Information.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eThe final codebook contained 561 codes, ranging from 83 to 102 codes per focus group. The seven domains of the NASSS framework with the emerging subcategories and descriptions are summarized in Table\u0026nbsp;2. Each category of the NASSS was classified as a \u003cem\u003efacilitator or barrier\u003c/em\u003e when at least two-thirds of the codes reflected a positive or negative influence on AI technology adoption, respectively. In cases when less than two-thirds of the codes reflected an overall positive or negative influence, the category was classified as \u003cem\u003emixed\u003c/em\u003e. The number of codes for each category and their valence (i.e., barrier, facilitator or mixed) can be seen in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n \u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n \u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;Table\u0026nbsp;2 The NASSS domains, categories, and descriptions regarding the adoption of AI technologies in psychotherapy\u003c/div\u003e\n \u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n \u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n \u003ctable id=\"Tabb\" style=\"width: 957px;\" border=\"1\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eNASSS Domain\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eCategory\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eDescription\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eCondition\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eThe nature and complexity of mental health conditions targeted by AI technologies.\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eType of mental health condition\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe specific mental health conditions that AI may target.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eSeverity of mental health condition\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe range of illness severity that AI may address.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eTechnology\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eThe characteristics of AI technologies.\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eUseful technology elements\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe specific elements that are considered useful for AI.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eCustomization to users\u0026rsquo; needs\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe extent to which AI can be tailored to specific users\u0026apos; needs.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eSupplementary use\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe use of AI as an adjunct rather than a replacement for traditional therapy.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eUsability for multiple user groups\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe extent to which AI is suited for use across different groups.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eUser-friendliness\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe ease and intuitiveness with which users can use AI.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eEvidence-based effectiveness\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eEvidence supporting AI\u0026rsquo;s effectiveness in improving mental health outcomes.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 413px;\" colspan=\"2\" align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eValue proposition\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eThe value of AI technologies for users.\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eIncreasing availability of access\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe ability for users to access AI at their convenience.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eImproved efficiency\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003ePotential improvements in efficiency offered by AI.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eShifts in depths of therapy elements\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe risk of AI oversimplifying therapy and reducing nuanced insights.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eAI as third element\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe role of AI as an additional agent during psychotherapy sessions.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003e(Commercial) conflict of interest\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe influence of commercial interests in the adoption of AI.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eAdopters\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eUser factors influencing the adoption of AI technologies.\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eAffinity for technology\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eUsers\u0026rsquo; comfort level and enthusiasm in using AI.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eTraining requirement\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe need for training to use AI effectively.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eLack of control\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe potential loss of control over therapy processes due to AI.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eTrust and openness\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe degree of trust in and openness for AI among users.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eTechnology understanding\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe knowledge or technical literacy required for effective use of AI.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eLack of human contact\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe absence of social and emotional connection after the integration of AI.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 930.051px;\" colspan=\"3\" align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eTable\u0026nbsp;2\u003c/strong\u003e (continued).\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eNASSS Domain\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eCategory\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eDescription\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eChanges in the importance of conventional psychotherapy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe potential of AI changing the perceived need for conventional therapy.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eTechnology dependency\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003e(Over-) Reliance on AI technologies.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003ePersistence of safety-seeking/ avoidance behavior\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe potential of AI to reinforce users\u0026apos; avoidance of difficult interactions or situations.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eOrganizations\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eThe readiness and capacity of an organization for AI technologies.\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eDifferences in the type of institution\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eUnderlying differences in the type of therapy-providing institutions influencing the use of AI.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eDifferences in therapy methods\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe compatibility of different therapeutic approaches with the potential use of AI.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eDigitalization\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe extent of digital infrastructure present in institutions.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003ePre-use testing opportunities\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe chance to try the AI at least once before use.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eOther incentives\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eExisting services or support structures necessary for adopting AI.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eResource and workload considerations\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe resources required for integrating AI.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eWider system\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eThe structure, dynamics, and capacity of the system in which AI technologies are introduced.\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eCost coverage\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe extent to which the use of AI is covered by insurance or other funding sources.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eEconomic feasibility\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eEconomic considerations regarding the adoption of AI.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eRegulations and regulatory environment\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eLegal and policy frameworks guiding the use of AI.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eData protection measures\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eMeasures ensuring data protection for the use of AI.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eLiability\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eDetermining accountability for AI outcomes.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eTransformations of existing structures\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eShifts in the structures of the mental healthcare system.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 413px;\" colspan=\"2\" align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eEmbedding and adaptation over time\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eThe adjustments in the adoption of AI technologies over time.\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eSpeed of implementation process\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe time it takes to integrate and adopt AI.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 92px;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd style=\"width: 321px;\" align=\"left\"\u003e\n \u003cp\u003eOngoing scientific validation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 517.051px;\" align=\"left\"\u003e\n \u003cp\u003eThe necessity of ongoing research to validate the efficacy of AI prior to its use.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eNote\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNumbers on the upper right corner of each category indicate the total number of codes assigned to that category. Tile size reflects the frequency of codes.\u003c/p\u003e\n\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n \u003ch2\u003e1. Condition\u003c/h2\u003e\n \u003cp\u003eFor the domain \u003cstrong\u003econdition\u003c/strong\u003e, two categories emerged as mixed factors (labeled M).\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e1.1. Type of mental health condition (M)\u003c/strong\u003e: Participants identified 15 mental health conditions that could potentially be addressed by AI technologies, with anxiety disorders and depression being the most frequently mentioned. Participants generally agreed on the potential usefulness of AI technologies for the following conditions: panic disorder, sleep disorder, hyperkinetic disorders, conduct disorders, substance use disorders, gender dysphoria, and borderline personality disorder. For instance, one patient explained:\u003c/p\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026ldquo;If there\u0026apos;s [\u0026hellip;] a button: \u0026apos;I\u0026apos;m having a panic attack. Help me\u0026apos;, and it tells me to do different exercises or whatever, I would take it.\u0026rdquo; (PA01).\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003cp\u003eOpinions were more divided for anxiety disorders, depression, eating disorders, ADHD/ADD, and suicide. For instance, therapists highlighted both potential benefits and risks of AI-based data tracking for eating disorders. Tracking physical activity and identifying deviations from usual activity patterns could provide valuable insights, as patients might not be fully aware of behaviors like excessive walking, which could contribute to difficulties in weight gain. However, such tracking could become a new obsession for patients, leading to an unhealthy focus on monitoring behaviors.\u003c/p\u003e\n \u003cp\u003eParticipants were skeptical about using AI technologies for conditions such as social phobia, psychosis, and paranoia. For instance, one therapist remarked that patients with paranoia or psychosis might be even more reluctant to adopt AI technologies, as these tools could amplify their existing suspicion and mistrust of their surroundings, leading to heightened discomfort and anxiety.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e1.2. Severity of mental health conditions (M)\u003c/strong\u003e: Participants emphasized the need to distinguish between mild, moderate, and severe mental health conditions to ensure appropriate adoption of AI. AI was unanimously deemed useful for milder conditions, but participants were hesitant about its application in more severe cases. They suggested a cautious, stepwise approach, starting with mild conditions and gradually expanding to more severe conditions.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n \u003ch2\u003e2. Technology\u003c/h2\u003e\n \u003cp\u003eIn the \u003cstrong\u003etechnology\u003c/strong\u003e domain, five categories were perceived as facilitators (labeled F) and one as mixed.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e2.1. Useful technology elements (F)\u003c/strong\u003e: Participants had various ideas for beneficial technology elements that would facilitate the adoption of AI technologies. Table\u0026nbsp;3 presents an overview of the designated technology features.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;Table\u0026nbsp;4\u0026nbsp;\u003cp\u003e\u003cem\u003eUseful technology elements retrieved from the focus group discussions\u003c/em\u003e\u003c/p\u003e\u0026nbsp;\u003c/div\u003e\n \u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n \u003ctable id=\"Tabc\" border=\"1\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eFeature\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eInformation purposes\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePsychoeducation\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGeneral information\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eEmotional and behavioral management\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eEmotion regulation\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMood tracking\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMood detection\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSkills\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRelaxation exercises\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eLogging and monitoring/ tracking\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTriggers\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDifferent types of diaries\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eActivity tracking\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRecording of critical events\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIdentifying patterns\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eTherapist support and process analysis\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eData analysis for therapists\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAnalysis of the therapy process\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMonitoring the development of the therapeutic relationship\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDecision support for treatment options\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDecision support for the therapeutic approach\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eSelf-management and organization\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHomework assignments\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReminders\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCreating a daily structure\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e(Self-help) Exercises\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eEmergency plan\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReinforcement plan/ system\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eSafety and risk management\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDetection of suicidality\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eTechnical and interactive features\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAvatar/ representation of the therapist\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAutomatic application filling\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eVoicemail\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eNote\u003c/em\u003e. \u003csup\u003ea\u003c/sup\u003e = not specified further\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003eThe most prominent features included psychoeducational content, mood tracking, different types of diaries, and self-help exercises like imaginary journeys. For instance, the value of mood tracking beyond fixed points was highlighted, especially when daily routines are disrupted, to provide deeper insights into a patient\u0026rsquo;s emotional state.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e2.2. Customization to users\u0026rsquo; needs (F)\u003c/strong\u003e: Another critical factor for the adoption of AI technologies in mental healthcare is its ability to be customized to users\u0026rsquo; individual needs. Personalization might help alleviate symptoms, while its absence might even worsen them:\u003c/p\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026ldquo;Every panic attack manifests itself differently, and what helps one person may not help another at all. You tell one person to concentrate on your breathing, and then they start to hyperventilate. You tell the other person to concentrate on your breathing and they calm down. [\u0026hellip;]\u0026rdquo; (PA03).\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\u003cstrong\u003e2.3. Supplementary use (F)\u003c/strong\u003e: Further, participants from both groups stressed using AI only as a supplement technology, not as a replacement for traditional therapy or a standalone solution, since they view human therapists as irreplaceable in the therapeutic process.\u003cbr\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e2.4. User-friendliness (F)\u003c/strong\u003e: Participants stressed that the technology should be easy to use, especially to reduce cognitive and emotional strain for individuals who may already feel overwhelmed.\u003c/p\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026ldquo;I think it should be very intuitive and not so complex, because especially if you have depression or something like that, you just don\u0026apos;t have the strength to get into it.\u0026rdquo; (PA08).\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cstrong\u003e2.5. Evidence-based effectiveness (F)\u003c/strong\u003e: Finally, participants emphasized the need for evidence supporting the effectiveness of AI technologies in improving outcomes.\u003c/p\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026ldquo;There should be studies that prove that AI-supported interventions have an effect, which is why it makes sense to use it, but the effect is not as great as psychotherapy from a real person.\u0026rdquo; (TH14).\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cstrong\u003e2.6. Usability for different user groups (M)\u003c/strong\u003e: Another AI adoption factor was its usability for multiple user groups. Three different sociodemographic areas were mentioned in which adoption barriers should be addressed: age, socioeconomic status, and culture and language:\u003c/p\u003e\n \u003cp\u003eFirst, older generations should be enabled to use emerging technologies.\u003c/p\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026ldquo;It is important to have a different perspective for children and young people than for adults or geriatric patients. I can possibly see difficulties there [\u0026hellip;], due to their probably rather limited use of digital devices. In other words, I ask myself, are these older patients then excluded? That should, of course, be avoided.\u0026rdquo; (TH18).\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003cp\u003eSecond, usage should be possible regardless of personal financial background, as participants emphasize potential inequalities for those with a lower socio-economic status.\u003c/p\u003e\n \u003cp\u003eAnd third, adaptations to different cultural and language backgrounds should be considered.\u003c/p\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026ldquo;[\u0026hellip;] I believe that there will also be a large section of the population that you do not deal with because it will then be very complicated to translate into simple language applications or actually offer something in their native language.\u0026rdquo; (TH11).\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n \u003ch2\u003e3. Value proposition\u003c/h2\u003e\n \u003cp\u003eIn the domain \u003cstrong\u003evalue proposition\u003c/strong\u003e, five categories emerged as either facilitators or barriers (labeled B) to AI adoption in mental healthcare.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e3.1. Increasing availability of access (F)\u003c/strong\u003e: A key benefit of AI technologies is accessibility, both in terms of location and timing. Participants highlighted the advantage of using AI flexibly in different places and its critical role in timely supporting patients during moments of crisis by providing immediate access to therapy resources.\u003c/p\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026ldquo;If I imagine: \u0026lsquo;Okay, I can fall back on previous therapy content, I\u0026apos;m in a crisis at the moment and can simply call it up\u0026rsquo;, I don\u0026apos;t necessarily need the therapist for that, that in turn can also have positive effects in terms of autonomy.\u0026rdquo; (TH18).\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cstrong\u003e3.2 Improved efficiency (F)\u003c/strong\u003e: AI could enhance efficiency by automating certain tasks, such as data analysis and administrative processes, enabling therapists to focus on patient care and reducing the strain caused by high patient volumes and long waitlists.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e3.3 Shifts in depths of therapy elements (B)\u003c/strong\u003e: Concerns were expressed that AI technologies might oversimplify complex therapeutic dynamics and reduce the value of nuanced observation. AI might overlook interpersonal dynamics and subtle behavioral cues, which are essential for forming accurate therapeutic judgments.\u003c/p\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026ldquo;[\u0026hellip;] I wonder if [\u0026hellip;] [AI is] too simplistic because we [humans] are actually much more complex than that, more than could be broken down into an algorithm that works when certain words are written, like in chatbots or similar things. I mean, the risk that we overlook something or miss something [\u0026hellip;] is very high.\u0026rdquo; (TH02).\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cstrong\u003e3.4 AI as a third element (B)\u003c/strong\u003e: Further, the potential drawback of AI as an additional agent in therapy was discussed, with concerns that it could disrupt the therapeutic relationship by causing confusion or conflicting interpretations, which can complicate the process and detract from the patient-therapist connection:\u003c/p\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026ldquo;There is then a third element that may influence the therapeutic relationship. [....] Then we have triangulation, [\u0026hellip;], which I think is complex. In principle, it\u0026apos;s a bit like two therapists working on the same patient.\u0026rdquo; (TH01).\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cstrong\u003e3.5 (Commercial) conflicts of interest (B)\u003c/strong\u003e: Another potential disadvantage might arise from commercial interests in adopting AI technologies. Concerns were expressed about profit-driven motives and exploitation of sensitive data for commercial gain, as technologies are rarely developed by non-profit companies.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\n \u003ch2\u003e4. Adopters\u003c/h2\u003e\n \u003cp\u003eAt the \u003cstrong\u003eadopter level\u003c/strong\u003e, nine distinct categories emerged as facilitators or barriers.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e4.1 Affinity for technology (F)\u003c/strong\u003e: Participants highlighted the importance of personal affinity for technology in its adoption, noting that individual user preferences affect how much they want to get involved in technology use and deal with its additional demands.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e4.2 Trust and openness (F)\u003c/strong\u003e: Trust in and openness towards AI is crucial, especially in intimate or emotionally charged contexts, given the sensitive nature of the topics addressed in psychotherapy and the data shared with these technologies.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e4.3 Training requirement (F)\u003c/strong\u003e: Since integrating new technologies can be time-consuming, therapists and patients suggested that training, through videos or practical workshops, could be an essential support mechanism. This was considered especially important in smaller practices or clinics where staff need to be trained quickly to avoid staff shortfall.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e4.5 Technology understanding (F)\u003c/strong\u003e: Closely related to the need for training is the level of users\u0026rsquo; knowledge or understanding to use AI technologies effectively. Therapists expressed the need to understand the technology to recommend it confidently to patients or to answer patients\u0026rsquo; questions. While patients found it essential that the therapists understand the technology, they did not emphasize the importance of their own understanding.\u003c/p\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026ldquo;[\u0026hellip;] Therapists [\u0026hellip;] [should] also [be] familiar with it. So that they can also give me recommendations or perhaps explain how to use it. And that they\u0026apos;re not just at the level of saying: \u0026apos;Well, there\u0026apos;s this app, you can use it, it\u0026apos;s good, but I don\u0026rsquo;t really know much else about it\u0026apos;.\u0026rdquo; (PA06).\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cstrong\u003e4.5 Lack of control (B)\u003c/strong\u003e: Therapists were concerned about the potential loss of control in the therapeutic process, such as their influence on the patient. Integrating AI into psychotherapy raised questions about maintaining control and ensuring it aligns with the therapist\u0026rsquo;s approach and methods.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e4.6 Lack of human contact (B)\u003c/strong\u003e: Another emerging concern for both user groups was the lack of human contact, especially the absence of social and emotional connection. The loss of human connection, which is generally perceived as a main mechanism in psychotherapy, may hinder patients\u0026rsquo; ability to open up and engage in meaningful and healing therapeutic work.\u003c/p\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026ldquo;So, I wouldn\u0026apos;t feel seen or acknowledged at all because I would just know: \u0026lsquo;Okay, it\u0026apos;s just an algorithm.\u0026rsquo; Artificial intelligence or not. What I would miss is the living counterpart, emphasis on living.\u0026rdquo; (PA03).\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cstrong\u003e4.7 Changes in the importance of conventional psychotherapy (B)\u003c/strong\u003e: Further, concerns were expressed that integrating AI could diminish the perceived value of and demand for in-person psychotherapy, leading to fewer therapy slots and steering patients away from traditional, personalized care.\u003c/p\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026quot;I can only put it plainly: there\u0026rsquo;s more of a concern that they\u0026rsquo;ll think, \u0026lsquo;Oh, great. A cheap, easy, and quickly available alternative - then we can cut costs elsewhere\u0026rsquo;.\u0026quot; (PA03).\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cstrong\u003e4.8 Technology dependency (B)\u003c/strong\u003e: Both therapists and patients were concerned about the potential for technology over-reliance. With its constant availability, there are fears that it could lead to an unhealthy dependence on devices rather than fostering real-world coping mechanisms.\u003c/p\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026quot;What I mentioned earlier as a pro argument could also be seen as a con argument here: the shift from independence from the therapist to dependence on a device.\u0026quot; (TH18).\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cstrong\u003e4.9 Persistence of safety-seeking and avoidance behavior (B)\u003c/strong\u003e: Finally, the use of such technologies might reinforce avoidance or safety behaviors in patients. The ease of using AI, especially for patients with social anxiety or phobias, could encourage avoidance of real-world challenges, hindering therapeutic improvements. While AI may offer immediate relief by eliminating the need for human interaction or enhancing perceived control through safety behaviors such as using conversational agents, it risks distracting from deeper emotional work.\u003c/p\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026ldquo;But it can perhaps also lead to patients starting to avoid contact or the relationship difficulties that only become apparent in therapy through shared experience, so that they don\u0026apos;t arise in the first place.\u0026rdquo; (TH02).\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\n \u003ch2\u003e5. Organizations\u003c/h2\u003e\n \u003cp\u003eAt the \u003cstrong\u003eorganizational\u003c/strong\u003e level, six categories emerged as either facilitator, barrier, or mixed factor.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e5.1 Pre-use testing opportunities (F)\u003c/strong\u003e: Opportunities to interact with and test AI technologies prior to their use are essential. Such opportunities can be created, for instance, through organizational support from the top down. This would allow individuals to test the technology in a neutral setting before using it with patients.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e5.2 Other incentives (F)\u003c/strong\u003e: Both patients and therapists believed that additional incentives should be introduced to support the integration of AI technologies into psychotherapy, such as a dedicated hotline for addressing questions that arise during and after the implementation process.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e5.3 Resource and workload considerations (B)\u003c/strong\u003e: Therapists expressed concerns that AI integration would initially create more effort than relief, with the required resources, particularly the time investment, frequently described as a barrier. Hence, they stressed the need for AI adoption to cause minimal additional workload, considering the already demanding nature of their profession.\u003c/p\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026ldquo;For me, [\u0026hellip;] this familiarization period would be very, very important, because I think it\u0026apos;s very important for us, but above all, for the patients, that you don\u0026apos;t somehow spend a whole hour of therapy on it [the familiarization period], so to speak, to deal with it first.\u0026rdquo; (TH17).\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cstrong\u003e5.4 Differences in therapy methods (M)\u003c/strong\u003e: Therapists usually specialize in a specific therapeutic approach, which may vary in compatibility with AI technologies. AI might be more easily integrated into structured approaches like cognitive behavioral therapy, where predefined methods and tasks are common. However, more dynamic and individual approaches, such as psychodynamic or systemic therapies, may face challenges since their change mechanisms rely on countertransference, childhood experiences, and deeper emotional engagement.\u003c/p\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026ldquo;I believe that behavioral therapy has far fewer problems than psychodynamic psychology in the whole complex that is being discussed here. And I think psychoanalysis is probably completely excluded.\u0026rdquo; (TH06).\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cstrong\u003e5.5 Differences in the type of institution (M)\u003c/strong\u003e: The type of institution providing therapy also impacts AI adoption, as therapists work in diverse settings from private practices to large clinics. Clinics could be more inclined to adopt AI technologies, as they tend to have the resources and infrastructure to integrate new technologies more easily. However, adoption decisions may also be influenced by financial constraints, administrative structures, and the overall attitude within the institution.\u003c/p\u003e\n \u003cp\u003e\u003cem\u003e\u0026ldquo;At least in my clinic I can imagine that the clinic management or the head physician could be interested in this, because they tend to have such a market-oriented awareness and say that innovations and things like that are good and can certainly be incorporated.\u0026rdquo; (TH11).\u003c/em\u003e\u003c/p\u003e\n \u003cp\u003eHowever, the downsides of AI adoption in clinics with top-down decision-making were also discussed:\u003c/p\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026ldquo;So, if I work in a clinic or in an outpatient clinic that I don\u0026apos;t manage myself, so to speak, then I\u0026apos;m dependent on the decisions of the person in charge. Whereas if I have my own outpatient practice, then I\u0026apos;m more independent and can introduce it more easily.\u0026rdquo; (TH18).\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cstrong\u003e5.6 Digitalization (M)\u003c/strong\u003e: Participants felt that digitalization in psychotherapy is not yet advanced enough for the introduction of AI tools. Concerns were raised about the slow digital transformation in healthcare, with many institutions still lacking basic infrastructure like Wi-Fi.\u003c/p\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026ldquo;I believe that we haven\u0026apos;t yet reached the point in the development of digitalization where it really saves time, where it [technology adoption] becomes automatic and where people have internalized it in such a way that they can say: \u0026lsquo;Okay, it [technology adoption] makes my work easier.\u0026rdquo; (TH09).\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\n \u003ch2\u003e6. Wider system\u003c/h2\u003e\n \u003cp\u003eAt the wider system level, six categories emerged as either facilitator, barrier, or mixed factor.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e6.1 Cost coverage (F)\u003c/strong\u003e: The prevailing topic was cost coverage. Therapists expressed the need for reimbursement strategies, and both user groups advocated for standardized insurance practices to ensure equitable access for patients.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e6.2 Economic feasibility (F)\u003c/strong\u003e: According to participants, the coverage of development and maintenance costs of AI systems must be ensured to facilitate their adoption. Further, it was noted that economic feasibility should not be pursued at the expense of quality and long-term value.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e6.3 Liability (B)\u003c/strong\u003e: The issue of liability for AI outcomes emerged as another concern, highlighted by a hypothetical scenario.\u003c/p\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026ldquo;Let\u0026apos;s say I have suicide prevention software, and it somehow works better, ten times better than a therapist [\u0026hellip;] [in detecting risks]. But still, sometimes people die by suicide. The question then is: Who is somehow liable? Who bears the risk?\u0026rdquo; (TH08).\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cstrong\u003e6.4 Data protection measures (B)\u003c/strong\u003e: Participants expressed significant concerns about data security, emphasizing risks related to data breaches, unclear data processing, and storage.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e6.5 Regulations and regulatory environment (M)\u003c/strong\u003e: Regulations to guide the use of AI in psychotherapy were deemed necessary. Participants emphasized the importance of clear legal frameworks, both nationally and internationally, through centralizing regulatory efforts to ensure data protection and quality standards and prevent misuse of sensitive information. However, some voiced concerns about overregulation hindering flexibility.\u003c/p\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026nbsp;\u0026ldquo;If this is always linked to any guidelines, then it may be that the whole thing becomes quite a cage as far as the application is concerned.\u0026rdquo; (TH09).\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cstrong\u003e6.6 Transformations of existing structures (M)\u003c/strong\u003e: Participants from both groups emphasized the need for a thoughtful approach between using AI to complement existing therapies and questioning whether its use would improve care or just address supply shortages short-term.\u003c/p\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026ldquo;But we could perhaps become even faster. We could become even more productive. Perhaps we could see even more patients. Treat even more. [...] Whether that is necessarily a good thing remains to be seen.\u0026rdquo; (TH10).\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\n \u003ch2\u003e7. Embedding and adaptation over time\u003c/h2\u003e\n \u003cp\u003eRegarding the embedding and potential adaptations over time, two categories were derived.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e7.1 Ongoing scientific validation (F)\u003c/strong\u003e: Participants highlighted the necessity of ongoing research to validate the efficacy of AI technologies. While initial studies exist, the consensus was that widespread implementation requires constant research, including pilot projects and rigorous testing to meet clinical standards.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e7.2 Speed of the implementation process (M)\u003c/strong\u003e: Opinions were mixed regarding the pace of AI integration into psychotherapy. Some participants acknowledged the unpredictability of technological advancements, suggesting that the adoption process could proceed rapidly. This is particularly true for simpler and more general AI technologies, such as those for text or image processing, which are projected to be implemented within the next three to five years. However, they anticipated that more specialized and niche applications, such as standalone AI-enabled technologies for VR applications in exposure therapy, take longer to implement, approximately ten years. Further, systemic challenges, including legislative and funding uncertainties, were highlighted as significant obstacles to AI implementation.\u003c/p\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026ldquo;This will probably take decades until Germany might actually reach the point where they say: \u0026lsquo;OK, AI-supported psychotherapy, [\u0026hellip;] maybe\u0026rsquo;. But only if all the framework conditions are 100% perfectly balanced and clarified.\u0026quot; (PA05).\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\n \u003ch2\u003e8. NASSS extension \u0026ndash; Time of use\u003c/h2\u003e\n \u003cp\u003eAs the aim of this research is to gain a comprehensive understanding of patients\u0026rsquo; and therapists\u0026rsquo; perspectives on the use of AI in psychotherapy, it is crucial to identify the stages and time points in the therapeutic process where technologies are perceived most beneficial. Therefore, we extend NASSS with \u0026ldquo;\u003cem\u003etime of use\u003c/em\u003e\u0026rdquo; as a new domain.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e8.1 Use in diagnostics and first screening (M)\u003c/strong\u003e: Participants acknowledged AI\u0026rsquo;s potential to conduct preliminary diagnostic evaluations by analyzing patient-completed questionnaires or assessments before therapist review. While therapists viewed it as useful to streamline workflows, concerns were raised about over-reliance, particularly for complex cases where personal history and symptom nuances may be overlooked.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e8.2 Use (during waiting times) before therapy (M)\u003c/strong\u003e: Participants saw the potential of AI to offer therapeutic resources during waiting periods or appointment delays, enabling early engagement. However, some emphasized that, especially in the beginning, therapeutic relationships should include personal interaction, which AI cannot replicate.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e8.3 Use between therapy sessions (F)\u003c/strong\u003e: Participants judged the use of AI technologies between therapy sessions positively, considering them as a tool to offer continuity and fill gaps in the process, such as providing tailored exercises.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e8.4 Use in aftercare (F)\u003c/strong\u003e: AI technologies could also serve as a valuable resource for post-therapy support, to maintain continuity, offer security, and ease the transition to self-management, particularly for patients feeling unprepared to navigate challenges independently. Participants highlighted its potential for relapse prevention and supporting the reduction of therapeutic contact.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eWith the rise of AI technologies and growing pressure on therapists and mental health systems, applying the NASSS framework to understand both patients\u0026rsquo; and therapists\u0026rsquo; perspectives offers valuable insights into the challenges and needs of AI adoption. Although patients and therapists represent distinct user groups with different backgrounds, their discussions revolved around largely overlapping topics, with clear differences in only two areas: only therapists highlighted the need to minimize AI-related workload due to limited time and resources, and only patients emphasized AI\u0026rsquo;s potential to improve efficiency. Notably, nearly half of all codes were allocated to the technology domain (such as its features and functionality) and the adopter domain (related to users\u0026rsquo; role), suggesting these areas are of great interest. The discussion highlights the most frequently cited categories, grouped into four broad areas:\u003c/p\u003e \u003cp\u003eFirst, the findings indicate that \u003cem\u003eheterogeneity in psychotherapy\u003c/em\u003e, including the specific \u003cem\u003etype of mental health condition\u003c/em\u003e, is an important factor influencing AI adoption. The positive attitudes toward AI technology for depression and anxiety disorders align with trends in research and technological development, which predominantly focus on these conditions (e.g., \u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e,\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e,\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e\u003c/sup\u003e). The reluctance to consider AI technologies for more complex mental illnesses, such as psychotic disorders, aligns with their broader hesitation toward their use in \u003cem\u003esevere symptomatology\u003c/em\u003e. In these cases, such as suicidality, hesitation was only alleviated when the AI was intended solely for supplementary support or early detection rather than for use during acute phases or as a standalone solution.\u003c/p\u003e \u003cp\u003eHeterogeneity at the organizational level, including different \u003cem\u003etherapeutic approaches\u003c/em\u003e, also needs to be considered. Consistent with our findings, prior research suggests that therapists' theoretical orientations influence their acceptance of new technologies, with psychoanalytic therapists showing lower adoption rates\u003csup\u003e\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e. Similarly, we found that AI integration is more compatible with CBT due to its structured, manualized, and rule-based methodologies. Another key barrier in this context is the variation in \u003cem\u003eorganizational and institutional environments\u003c/em\u003e therapists work in. Therapists noted that in clinical settings, decisions regarding technology adoption are often outside their control and depend on other stakeholders, such as clinic management. Participants believed that while clinicians and management are more motivated to adopt technologies, administrative stakeholders tend to hinder adoption due to investment costs. In contrast, therapists in private practice have greater autonomy but face the challenge of managing adoption on their own.\u003c/p\u003e \u003cp\u003eSecond, \u003cem\u003euser-specific human factors\u003c/em\u003e also played a pivotal role in AI adoption, with the \u003cem\u003elack of human contact\u003c/em\u003e emerging as an especially critical aspect. This supports previous research emphasizing the need to preserve the human element in patient-therapist interaction when adopting AI technologies, especially in the emotionally sensitive field of mental healthcare\u003csup\u003e\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e,\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e. Such concerns stem from fears that AI could undermine authentic human interaction, potentially disrupting meaningful therapeutic relationships\u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e,\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e,\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u003c/sup\u003e. For instance, meaningful relationships are shaped by emotional shifts, like countertransference in psychodynamic psychotherapy. Therapists were hesitant that AI technologies could address these shifts appropriately, stressing the need for human responsiveness. The lack of this human element may hinder therapeutic progress, given the well-established importance of the therapeutic relationship in psychotherapy outcomes\u003csup\u003e\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e. Despite the growing body of research on AI systems establishing a human bond and their promising advancements\u003csup\u003e\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e\u003c/sup\u003e, such as enhancing social presence in chatbot-based interactions to foster rapport\u003csup\u003e\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e\u003c/sup\u003e, our findings underscore the continued importance that patients and therapists place on human contact. Furthermore, AI\u0026rsquo;s role as a \u003cem\u003esupplementary tool\u003c/em\u003e emerged as a key facilitator. Participants agreed that human therapists are irreplaceable in the therapeutic process, and that AI technologies, such as AI-enabled mental health apps suggesting interventions, should serve only as a supplement, aligning with existing literature\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e,\u003cspan additionalcitationids=\"CR54\" citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eMoreover, \u003cem\u003ecustomization to users\u0026rsquo; needs\u003c/em\u003e is one strategy that can help address concerns about the lack of human connection. In our study, customization includes adapting to patients\u0026rsquo; therapy experience and knowledge, which helps avoid repetition and tailors advice to prevent suggestions that may worsen the situation. This personalized approach might foster a sense of connection and relevance, thereby mitigating concerns that technology may hinder the human element of therapy. The importance of customization is reflected in our findings, as users viewed customization as a critical feature, in line with previous research\u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e. While the literature often focuses on technology customization for patients, for instance, to strengthen the therapeutic relationship by personalizing the therapy style and avatar of chatbots\u003csup\u003e\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e\u003c/sup\u003e, less attention is given to therapists\u0026rsquo; perspectives. Our therapists stressed that customization on their end involves aligning technologies also with their practice, ensuring it is tailored to specific tasks that support their work and maintain authenticity, without compromising their therapeutic framework. Ideally, tailoring the technology to each user's needs enhances its \u003cem\u003eusability across different user groups\u003c/em\u003e, a key facilitator highlighted in the literature, particularly regarding age\u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e,\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e\u003c/sup\u003e. Studies have found hesitations towards AI technologies in older adults\u003csup\u003e\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e\u003c/sup\u003e, while younger patients tend to show greater engagement with technologies\u003csup\u003e\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e. However, our findings even extend this by identifying additional sociodemographic factors, such as the patient\u0026rsquo;s financial situation, cultural background, and language requirements.\u003c/p\u003e \u003cp\u003eThird, the \u003cem\u003eaccessibility of AI technologies\u003c/em\u003e is also a crucial consideration. Participants emphasized the benefit of continuous support and \u003cem\u003eflexible access\u003c/em\u003e to therapy content, particularly across locations and times. Other research also underscores the positive perception of accessible AI technologies, such as those via smartphones, which can reach patients in high-risk situations\u003csup\u003e\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e\u003c/sup\u003e and offer 24/7 support\u003csup\u003e\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e\u003c/sup\u003e. However, in our study, participants had a more nuanced view: they noted that constant accessibility could lead to \u003cem\u003edependency\u003c/em\u003e, making the absence of AI technology feel unsettling. This dependency could even exacerbate symptoms by fostering compulsive behaviors, such as feeling compelled to engage excessively with technology. As existing research primarily focuses on the harm of dependency on chatbots\u003csup\u003e\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e\u003c/sup\u003e, our findings extend this concern to AI applications in psychotherapy in general, emphasizing the broader risk. Additionally, both user groups stressed the risk that AI technologies, through individualized and non-confrontational interactions, may reinforce \u003cem\u003esafety-seeking or avoidance behavior\u003c/em\u003e by reducing the exposure to therapeutically relevant challenges. In psychotherapy, such challenges \u0026ndash; like deliberate ambiguous or paradoxical therapist responses \u0026ndash; are often critical for emotional processing and behavioral change. This concern was especially noted in relation to anxiety disorders, particularly social anxiety. While initially, since it has been shown that individuals with social anxiety tend to disclose more information to conversational agents than to human counterparts\u003csup\u003e\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e,\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e\u003c/sup\u003e, this might seem beneficial by promoting communication and engagement, it might hinder therapeutic process by limiting therapeutic opportunities to engage in real social interactions\u003csup\u003e\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eFourth, different \u003cem\u003estructures surrounding AI adoption\u003c/em\u003e also play a pivotal role. Having a robust \u003cem\u003edigital infrastructure\u003c/em\u003e is the foundation of AI adoption, however, the current infrastructure in both clinical and private practice settings remains insufficient. While the importance of adequate digital infrastructure in healthcare is widely recognized\u003csup\u003e\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e\u003c/sup\u003e, in practice, Germany lags behind global leaders in the development of AI technologies for mental health disorders\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e and other European countries in the digitalization of medical services\u003csup\u003e\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e\u003c/sup\u003e. This lack of favorable conditions compounds another challenge identified by therapists: \u003cem\u003eresource and workload constraints\u003c/em\u003e. Adopting new technologies requires time and resources, and both our findings and prior studies indicate that therapists are rather reluctant to engage with time-intensive digital implementations\u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e. Dealing with poor digital infrastructure adds workload and complicates the AI adoption process. It also undermines other facilitators, such as \u003cem\u003epre-use testing opportunities\u003c/em\u003e and comprehensive \u003cem\u003etraining\u003c/em\u003e, that have been linked to facilitating adoption\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e,\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e,\u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eFurther, financial and structural transformations at the wider-system level, such as \u003cem\u003ecost coverage\u003c/em\u003e, could help overcome technology and adopter-level barriers, for instance by ensuring accessibility for diverse user groups. Cost coverage by health insurance companies is a prerequisite for patients to adopt AI technologies\u003csup\u003e\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e. Additionally, therapists noted the need for affordable technologies and reimbursement\u003csup\u003e\u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e69\u003c/span\u003e\u003c/sup\u003e. However, participants indicated that meeting cost coverage needs might necessitate a restructuring of current reimbursement mechanisms. Moreover, participants anticipated that system-level structures, such as patient referral pathways, would undergo a \u003cem\u003etransformation\u003c/em\u003e, as AI could increase access to therapy, allowing for a larger number of patients to be treated. This, in turn, raised some concerns, especially regarding the treatment quality, with the risk that these changes might offer only a temporary increase in access without addressing the underlying systemic issues, making them a \u0026ldquo;quick fix\u0026rdquo; rather than a lasting solution, and psychotherapy becoming less patient-centered and rather driven by economic efficiency.\u003c/p\u003e \u003cp\u003eFinally, technology \u003cem\u003eembedding and adaptation over time\u003c/em\u003e have been largely overlooked in existing literature\u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e,\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e. The majority of our participants expressed uncertainty about the speed of the adoption process and refrained from making generalizations across AI technologies. Instead, they anticipate slower adoption for more specialized and niche applications, while simpler and more general AI technologies are believed to be implemented quicker.\u003c/p\u003e \u003cp\u003eOur findings highlight some important implications for the development, design, and implementation of AI technologies and further research. First, addressing the specific technological features valued by patients and therapists can support adoption\u003csup\u003e\u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e\u003c/sup\u003e. Particularly, features enhancing patient autonomy, such as psychoeducational content to aid patients in understanding their conditions, self-help exercises like meditation or guided imagery, and diaries, are deemed most useful. Notably, a common characteristic of these features is their utility when therapists are unavailable and during pre-therapy phases, including diagnostic and initial screening, or while awaiting treatment. Second, aligning the purpose of the AI technology with users\u0026rsquo; preferred timing of use is critical to reduce hesitation. Participants in our study identified four key timepoints where they believed that AI could support and enhance the therapeutic process. At each of these stages \u0026ndash; diagnostics and screening, before therapy, between therapy sessions, and in aftercare \u0026ndash; AI\u0026rsquo;s use would need to be tailored to the specific needs and goals of that phase to be effective and well-received. Third, there was a substantial overlap in the topics discussed by patients and therapists. While addressing the distinct needs of each user group is essential for AI adoption, targeting shared barriers may simplify adoption and reduce complexity for both. Finally, our findings of mixed categories where the same aspect can be perceived as a barrier and facilitator in almost all NASSS domains highlight the complexity on every level of technology adoption. Understanding these dual-role factors is crucial for developing strategies that maximize their facilitative potential while minimizing drawbacks. For instance, the need for clear legal regulations for technology adoption was emphasized among participants and is also highlighted in the literature\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e,\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e. Regulations should clarify the use and storage of sensitive data, delineate tasks appropriate for AI technologies against those that must be performed by humans, and determine whether the use of AI technology should be mandatory or integrated in treatment guidelines. However, regulation should not limit therapists\u0026rsquo; treatment flexibility and hinder implementation. Future research is needed to better understand how regulatory frameworks can support the integration of AI while preserving ethical standards and the clinician\u0026rsquo;s role in therapy. Balancing these concerns is therefore key for fostering both ethical safeguards and innovation.\u003c/p\u003e \u003cp\u003eThis study has several limitations: First, the sample was imbalanced, with three focus groups with therapists and only two with patients. This occurred because participant recruitment followed the principle of data saturation\u003csup\u003e\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e\u003c/sup\u003e, meaning that data collection continued until no new topics emerged. The therapist focus groups were conducted first due to earlier sign-ups resulting in fewer patient focus groups, as these started later in the process, and saturation was achieved after five focus groups. Second, our relatively young study sample may reflect a more technology-engaged group\u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e,\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e, potentially skewing perceptions. Third, our study only included participants from across Germany; hence, their experiences and perceptions regarding the digital infrastructure and healthcare systems may not be generalizable to other countries. Healthcare policies and technological advancements differ widely across regions, with countries like the U.S. and China leading in AI development for mental healthcare\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. These regional variances in healthcare systems and regulations may influence how AI technologies are perceived and adopted into mental health treatment, limiting the generalizability of our findings to other contexts with different healthcare frameworks.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eOverall, the study highlights that AI adoption in mental healthcare is a complex and multifaceted issue, influenced by the subjectivity and heterogeneity of mental illnesses, and the diverse therapeutic approaches and workplace environments. Participants\u0026rsquo; needs, viewed as facilitating factors, emerged in every domain of the technology adoption process, and when appropriately leveraged, can ease AI adoption. However, given the general low use intentions of AI in mental healthcare, it is essential to design AI solutions that not only align with user needs but also alleviate their concerns. Addressing these barriers proactively holds the potential to overcome hesitation, realize the full potential of AI technology, and consequently, improve mental healthcare by mitigating the treatment gap.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e \u003ch2\u003eCompeting interests:\u003c/h2\u003e \u003cp\u003eAll authors declare no financial or non-financial competing interests.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eFunding:\u003c/h2\u003e \u003cp\u003eThis study was funded by a grant from the Volkswagen Foundation (Grant #: 98525). The funder played no role in study design, data collection, analysis, and interpretation of data, or the writing of this manuscript.\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eJ.C., S.G., and A.-K.K. conceptualized and design the study. J.C. and D.E. recruited participants and collected the data in the focus groups. J.C., D.E. and I.S. analyzed the focus group transcripts. J.C. wrote the original draft of the manuscript. E.L., D.E., I.S., S.G., and A.-K. K. contributed to the writing, interpretation, and editing of the manuscript. All authors read and approved the final manuscript.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe discussion guide, COREQ checklist and the description of AI and AI-enabled technologies in psychotherapy given to participants prior to the study is available on OSF (https://osf.io/9rt52/?view_only=a1a4b6a50d8f4d62b3b685eee6651a9b). The codebook and exemplary codes supplementing the study results are available in the online Supplementary Information.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eQin, X. \u0026amp; Hsieh, C.-R. Understanding and Addressing the Treatment Gap in Mental Healthcare: Economic Perspectives and Evidence From China. Inquiry. 57, 0046958020950566 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRoberts, T. \u003cem\u003eet al.\u003c/em\u003e Reconceptualising the treatment gap for common mental disorders: a fork in the road for global mental health? Br. J. Psychiatry 221, 553\u0026ndash;557 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWorld Health Organization. \u003cem\u003eWorld Mental Health Report: Transforming mental health for all\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.who.int/publications/i/item/9789240049338\u003c/span\u003e\u003cspan address=\"https://www.who.int/publications/i/item/9789240049338\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDeutsches \u0026Auml;rzteblatt. \u003cem\u003ePsychotherapie: L\u0026auml;ngere Wartezeiten auf Therapieplatz seit Strukturreform [Psychotherapy: Longer waiting times for therapy places since structural reform]\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.aerzteblatt.de/news/psychotherapie-laengere-wartezeiten-auf-therapieplatz-seit-strukturreform-cbef13cc-5ae\u003c/span\u003e\u003cspan address=\"https://www.aerzteblatt.de/news/psychotherapie-laengere-wartezeiten-auf-therapieplatz-seit-strukturreform-cbef13cc-5ae\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e1-4ae6-bf04-58bf00edee58 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSubotic-Kerry, M. \u003cem\u003eet al.\u003c/em\u003e While they wait: A cross-sectional survey on wait times for mental health treatment for anxiety and depression for Australian adolescents. BMJ open. 15, e087342 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWainberg, M. L. \u003cem\u003eet al.\u003c/em\u003e Challenges and Opportunities in Global Mental Health: a Research-to-Practice Perspective. Curr. Psychiatry Rep. 19, 28 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKleine, A.-K. \u003cem\u003eet al.\u003c/em\u003e AI-enabled clinical decision support tools for mental healthcare: A product review. Artif. Intell. Med. 160, 103052 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKleine, A.-K., Lermer, E., Cecil, J., Heinrich, A. \u0026amp; Gaube, S. Advancing mental health care with AI-enabled precision psychiatry tools: A patent review. Comput. Hum. Behav. Rep. 12, 100322 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOlawade, D. B. \u003cem\u003eet al.\u003c/em\u003e Enhancing mental health with Artificial Intelligence: Current trends and future prospects. J. Med. Surg. Public Health 3, 100099 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCecil, J., Kleine, A.-K., Lermer, E. \u0026amp; Gaube, S. Mental health practitioners\u0026rsquo; perceptions and adoption intentions of AI-enabled technologies: an international mixed-methods study. BMC Health Serv. Res. 25, 556 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBeg, M. J., Verma, M., M., V. C. K. M. \u0026amp; Verma, M. K. Artificial Intelligence for Psychotherapy: A Review of the Current State and Future Directions. Indian J. Psychol. Med. (2024) doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1177/02537176241260819\u003c/span\u003e\u003cspan address=\"10.1177/02537176241260819\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStade, E. C. \u003cem\u003eet al.\u003c/em\u003e Large language models could change the future of behavioral healthcare: a proposal for responsible development and evaluation. Npj Ment. Health Res. 3, 12 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSquires, M. \u003cem\u003eet al.\u003c/em\u003e Deep learning and machine learning in psychiatry: a survey of current progress in depression detection, diagnosis and treatment. Brain Inform. 10, 10 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRollwage, M. \u003cem\u003eet al.\u003c/em\u003e Using Conversational AI to Facilitate Mental Health Assessments and Improve Clinical Efficiency Within Psychotherapy Services: Real-World Observational Study. JMIR AI 2, e44358 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKoh, J., Tng, G. Y. Q. \u0026amp; Hartanto, A. Potential and Pitfalls of Mobile Mental Health Apps in Traditional Treatment: An Umbrella Review. J. Pers. Med. 12, 1376 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKleine, A.-K., Kokje, E., Lermer, E. \u0026amp; Gaube, S. Attitudes Toward the Adoption of 2 Artificial Intelligence\u0026ndash;Enabled Mental Health Tools Among Prospective Psychotherapists: Cross-sectional Study. JMIR Hum. Factors 10, e46859 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGutierrez, G., Stephenson, C., Eadie, J., Asadpour, K. \u0026amp; Alavi, N. Examining the role of AI technology in online mental healthcare: opportunities, challenges, and implications, a mixed-methods review. Front. Psychiatry 15, 1356773 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBeilharz, F., Sukunesan, S., Rossell, S. L., Kulkarni, J. \u0026amp; Sharp, G. Development of a Positive Body Image Chatbot (KIT) With Young People and Parents/Carers: Qualitative Focus Group Study. J. Med. Internet Res. 23, e27807 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBraun, P. \u003cem\u003eet al.\u003c/em\u003e Investigating information needs and preferences regarding digital mental health services among medical and psychology students in Germany: A qualitative study. Digit. Health 9, 20552076231173568 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCreed, T. A. \u003cem\u003eet al.\u003c/em\u003e Knowledge and Attitudes Toward an Artificial Intelligence-Based Fidelity Measurement in Community Cognitive Behavioral Therapy Supervision. Adm. Policy Ment. Health Ment. Health Serv. Res. 49, 343\u0026ndash;356 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGoodspeed, A. \u003cem\u003eet al.\u003c/em\u003e Leveraging the utility of pharmacogenomics in psychiatry through clinical decision support: a focus group study. Ann. Gen. Psychiatry 18, 13 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHenshall, C. \u003cem\u003eet al.\u003c/em\u003e Implementing a digital clinical decision support tool for side effects of antipsychotics: a focus group study. BMJ Ment Health 22, 56\u0026ndash;60 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRadovic, A. \u003cem\u003eet al.\u003c/em\u003e Perceived Facilitators of and Barriers to Implementation of a Decision Support Tool for Adolescent Depression and Suicidality Screening: Focus Group and Interview Study. JMIR Ment. Health 8, e26035 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGreenhalgh, T. \u003cem\u003eet al.\u003c/em\u003e The NASSS-CAT Tools for Understanding, Guiding, Monitoring, and Researching Technology Implementation Projects in Health and Social Care: Protocol for an Evaluation Study in Real-World Settings. JMIR Res. Protoc. 9, e16861 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJames, H. M., Papoutsi, C., Wherton, J., Greenhalgh, T. \u0026amp; Shaw, S. E. Spread, Scale-up, and Sustainability of Video Consulting in Health Care: Systematic Review and Synthesis Guided by the NASSS Framework. J. Med. Internet Res. 23, e23775 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBrown, P. \u003cem\u003eet al.\u003c/em\u003e Automated Virtual Reality Cognitive Therapy (gameChange) in Inpatient Psychiatric Wards: Qualitative Study of Staff and Patient Views Using an Implementation Framework. JMIR Form. Res. 6, e34225 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBerardi, C. \u003cem\u003eet al.\u003c/em\u003e Barriers and facilitators to the implementation of digital technologies in mental health systems: a qualitative systematic review to inform a policy framework. BMC Health Serv. Res. 24, 243 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlqahtani, F., Winn, A. \u0026amp; Orji, R. Co-Designing a Mobile App to Improve Mental Health and Well-Being: Focus Group Study. JMIR Form. Res. 5, e18172 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTerra, M., Baklola, M., Ali, S. \u0026amp; El-Bastawisy, K. Opportunities, applications, challenges and ethical implications of artificial intelligence in psychiatry: a narrative review. Egypt. J. Neurol. Psychiatry Neurosurg. 59, 80 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFreitag, B., Uncovska, M., Meister, S., Prinz, C. \u0026amp; Fehring, L. Cost-effectiveness analysis of mHealth applications for depression in Germany using a Markov cohort simulation. Npj Digit. Med. 7, 1\u0026ndash;13 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAafjes-van Doorn, K. Feasibility of artificial intelligence-based measurement in psychotherapy practice: Patients\u0026rsquo; and clinicians\u0026rsquo; perspectives. Couns. Psychother. Res. 25, e12800 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMendes-Santos, C., Nunes, F., Weiderpass, E., Santana, R. \u0026amp; Andersson, G. Understanding Mental Health Professionals\u0026rsquo; Perspectives and Practices Regarding the Implementation of Digital Mental Health: Qualitative Study. JMIR Form. Res. 6, e32558 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHogg, H. D. J. \u003cem\u003eet al.\u003c/em\u003e Stakeholder Perspectives of Clinical Artificial Intelligence Implementation: Systematic Review of Qualitative Evidence. J. Med. Internet Res. 25, e39742 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRichardson, J. P. \u003cem\u003eet al.\u003c/em\u003e Patient apprehensions about the use of artificial intelligence in healthcare. Npj Digit. Med. 4, 1\u0026ndash;6 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAbell, B. \u003cem\u003eet al.\u003c/em\u003e Identifying barriers and facilitators to successful implementation of computerized clinical decision support systems in hospitals: a NASSS framework-informed scoping review. Implement. Sci. 18, 32 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAhmed, M. I. \u003cem\u003eet al.\u003c/em\u003e A Systematic Review of the Barriers to the Implementation of Artificial Intelligence in Healthcare. Cureus 15, e46454 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBastoni, S. \u003cem\u003eet al.\u003c/em\u003e Factors Influencing Implementation of eHealth Technologies to Support Informal Dementia Care: Umbrella Review. JMIR Aging 4, e30841 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHummelsberger, P. \u003cem\u003eet al.\u003c/em\u003e Insights on the Current State and Future Outlook of AI in Health Care: Expert Interview Study. JMIR AI 2, e47353 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLim, W. M. What Is Qualitative Research? An Overview and Guidelines. Australas. Mark. J. (2024) doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1177/14413582241264619\u003c/span\u003e\u003cspan address=\"10.1177/14413582241264619\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTong, A., Sainsbury, P. \u0026amp; Craig, J. Consolidated criteria for reporting qualitative research (COREQ): a 32-item checklist for interviews and focus groups. Int. J. Qual. Health Care 19, 349\u0026ndash;357 (2007).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGBD 2019 Mental Disorders Collaborators. Global, regional, and national burden of 12 mental disorders in 204 countries and territories, 1990\u0026ndash;2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet Psychiatry 9, 137\u0026ndash;150 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFreeman, T. \u0026lsquo;Best practice\u0026rsquo; in focus group research: making sense of different views. J. Adv. Nurs. 56, 491\u0026ndash;497 (2006).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFereday, J. \u0026amp; Muir-Cochrane, E. Demonstrating Rigor Using Thematic Analysis: A Hybrid Approach of Inductive and Deductive Coding and Theme Development. Int. J. Qual. Methods 5, 80\u0026ndash;92 (2006).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcHugh, M. L. Interrater reliability: the kappa statistic. Biochem. Medica 276\u0026ndash;282 (2012) doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.11613/BM.2012.031\u003c/span\u003e\u003cspan address=\"10.11613/BM.2012.031\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHerbener, A. B., Klincewicz, M. \u0026amp; Damholdt, M. F. A narrative review of the active ingredients in psychotherapy delivered by conversational agents. Comput. Hum. Behav. Rep. 14, 100401 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBourla, A. \u003cem\u003eet al.\u003c/em\u003e Psychiatrists\u0026rsquo; Attitudes Toward Disruptive New Technologies: Mixed-Methods Study. JMIR Ment. Health 5, e10240 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBlease, C. \u003cem\u003eet al.\u003c/em\u003e Artificial Intelligence and the Future of Primary Care: Exploratory Qualitative Study of UK General Practitioners\u0026rsquo; Views. J. Med. Internet Res. 21, e12802 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHaber, Y., Levkovich, I., Hadar-Shoval, D. \u0026amp; Elyoseph, Z. The Artificial Third: A Broad View of the Effects of Introducing Generative Artificial Intelligence on Psychotherapy. JMIR Ment. Health 11, e54781\u0026ndash;e54781 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRuse, J. N., Schraube, E. \u0026amp; Rhodes, P. Left to their own devices: the significance of mental health apps on the construction of therapy and care. Subjectivity 31, 410\u0026ndash;428 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLambert, M. J. \u0026amp; Barley, D. E. Research summary on the therapeutic relationship and psychotherapy outcome. Psychother. Theory Res. Pract. Train. 38, 357\u0026ndash;361 (2001).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMorrow, E. \u003cem\u003eet al.\u003c/em\u003e Artificial intelligence technologies and compassion in healthcare: A systematic scoping review. Front. Psychol. 13, 971044 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee, J., Lee, D. \u0026amp; Lee, J. Influence of Rapport and Social Presence with an AI Psychotherapy Chatbot on Users\u0026rsquo; Self-Disclosure. Int. J. Human\u0026ndash;Computer Interact. 40, 1620\u0026ndash;1631 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBabu, A. \u0026amp; Joseph, A. P. Artificial intelligence in mental healthcare: transformative potential vs. the necessity of human interaction. Front. Psychol. 15, 1378904 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eG\u0026ouml;tzl, C. \u003cem\u003eet al.\u003c/em\u003e Artificial intelligence-informed mobile mental health apps for young people: a mixed-methods approach on users\u0026rsquo; and stakeholders\u0026rsquo; perspectives. Child Adolesc. Psychiatry Ment. Health 16, 86 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRebelo, A. D., Verboom, D. E., Dos Santos, N. R. \u0026amp; De Graaf, J. W. The impact of artificial intelligence on the tasks of mental healthcare workers: A scoping review. Comput. Hum. Behav. Artif. Hum. 1, 100008 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVossen, W., Szymanski, M. \u0026amp; Verbert, K. The effect of personalizing a psychotherapy conversational agent on therapeutic bond and usage intentions. in \u003cem\u003eProceedings of the 29th International Conference on Intelligent User Interfaces\u003c/em\u003e 761\u0026ndash;771 (Association for Computing Machinery, New York, NY, USA, 2024). doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1145/3640543.3645195\u003c/span\u003e\u003cspan address=\"10.1145/3640543.3645195\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBucci, S. \u003cem\u003eet al.\u003c/em\u003e \u0026ldquo;They Are Not Hard-to-Reach Clients. We Have Just Got Hard-to-Reach Services.\u0026rdquo; Staff Views of Digital Health Tools in Specialist Mental Health Services. Front. Psychiatry 10, (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFritsch, S. J. \u003cem\u003eet al.\u003c/em\u003e Attitudes and perception of artificial intelligence in healthcare: A cross-sectional survey among patients. Digit. Health 8, 20552076221116772 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eConnolly, S. L. \u003cem\u003eet al.\u003c/em\u003e Patient and provider predictors of telemental health use prior to and during the COVID-19 pandemic within the Department of Veterans Affairs. Am. Psychol. 77, 249\u0026ndash;261 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAktan, M. E., Turhan, Z. \u0026amp; Dolu, İ. Attitudes and perspectives towards the preferences for artificial intelligence in psychotherapy. Comput. Hum. Behav. 133, 107273 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLaestadius, L., Bishop, A., Gonzalez, M., Illenč\u0026iacute;k, D. \u0026amp; Campos-Castillo, C. Too human and not human enough: A grounded theory analysis of mental health harms from emotional dependence on the social chatbot Replika. New Media Soc. 26, 5923\u0026ndash;5941 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHu, B. How social anxiety leads to problematic use of conversational AI: The roles of loneliness, rumination, and mind perception. Comput. Hum. Behav. 145, 107760 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKang, S. \u0026amp; Gratch, J. Virtual humans elicit socially anxious interactants\u0026rsquo; verbal self-disclosure. Comput. Animat. Virtual Worlds 21, 473\u0026ndash;482 (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRodebaugh, T. L., Holaway, R. M. \u0026amp; Heimberg, R. G. The treatment of social anxiety disorder. Clin. Psychol. Rev. 24, 883\u0026ndash;908 (2004).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHassan, M., Kushniruk, A. \u0026amp; Borycki, E. Barriers to and Facilitators of Artificial Intelligence Adoption in Health Care: Scoping Review. JMIR Hum. Factors 11, e48633 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMajcherek, D., Hegerty, S. W., Kowalski, A. M., Lewandowska, M. S. \u0026amp; Dikova, D. Opportunities for healthcare digitalization in Europe: Comparative analysis of inequalities in access to medical services. Health Policy 139, 104950 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGolden, G. \u003cem\u003eet al.\u003c/em\u003e Applying artificial intelligence to clinical decision support in mental health: What have we learned? Health Policy Technol. 13, 100844 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLambert, S. I. \u003cem\u003eet al.\u003c/em\u003e An integrative review on the acceptance of artificial intelligence among healthcare professionals in hospitals. Npj Digit. Med. 6, 1\u0026ndash;14 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee, E. E. \u003cem\u003eet al.\u003c/em\u003e Artificial Intelligence for Mental Health Care: Clinical Applications, Barriers, Facilitators, and Artificial Wisdom. Biol. Psychiatry Cogn. Neurosci. Neuroimaging 6, 856\u0026ndash;864 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchueller, S. M., Neary, M., O\u0026rsquo;Loughlin, K. \u0026amp; Adkins, E. C. Discovery of and Interest in Health Apps Among Those With Mental Health Needs: Survey and Focus Group Study. J. Med. Internet Res. 20, e10141 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSaunders, B. \u003cem\u003eet al.\u003c/em\u003e Saturation in qualitative research: exploring its conceptualization and operationalization. Qual. Quant. 52, 1893\u0026ndash;1907 (2018).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"npj-mental-health-research","isNatureJournal":false,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"npjmentalhealth","sideBox":"Learn more about [npj Mental Health Research](https://www.nature.com/npjmentalhealth/)","snPcode":"44184","submissionUrl":"https://mts-npjmentalhealth.nature.com/cgi-bin/main.p...","title":"npj Mental Health Research","twitterHandle":"@npjmentalhealth\n","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"npj","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Mental healthcare, Artificial intelligence, Technology Adoption","lastPublishedDoi":"10.21203/rs.3.rs-6546781/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6546781/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eArtificial intelligence (AI) technologies in mental healthcare offer promising opportunities to reduce therapists\u0026rsquo; burden and enhance healthcare delivery, yet adoption remains challenging. This study identified key facilitators and barriers to AI adoption in mental healthcare, precisely psychotherapy, by conducting six online focus groups with patients and therapists, using a semi-structured guide based on the NASSS (Nonadoption, Abandonment, Scale-up, Spread, and Sustainability) framework. Data from \u003cem\u003eN\u003c/em\u003e\u0026thinsp;=\u0026thinsp;32 participants were analyzed using a combined deductive and inductive thematic analysis. Across the seven NASSS domains, 36 categories emerged. Sixteen categories were identified as factors facilitating adoption, including useful technology elements, the customization to user needs, and cost coverage. Eleven categories were perceived as barriers to adoption, encompassing the lack of human contact, resource constraints, and AI dependency. Further nine, such as therapeutic approach and institutional differences, acted as both facilitators and barriers depending on the context. Our findings highlight the complexity of AI adoption in mental healthcare and emphasize the importance of addressing barriers early in the development of AI technologies.\u003c/p\u003e","manuscriptTitle":"Navigating the complexity of AI adoption in psychotherapy by identifying key facilitators and barriers","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-05 11:11:07","doi":"10.21203/rs.3.rs-6546781/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-12-07T16:12:26+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-12-05T20:46:42+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"96405031097917174624427984174302026302","date":"2025-11-14T16:43:01+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"108065720816148235675109985433762831638","date":"2025-11-14T01:12:20+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"11782994364217690026775372826327023950","date":"2025-11-10T15:55:37+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"130356650730101948512045346655239319774","date":"2025-09-29T16:53:51+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"101543008720902920926491308922870779095","date":"2025-08-03T23:35:48+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"2144249969345652113047011256971394571","date":"2025-08-03T00:03:46+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-06-22T12:11:06+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"230113134442290680359031028682759315956","date":"2025-06-01T15:59:08+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-06-01T15:55:47+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-05-12T17:36:46+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-04-29T17:17:48+00:00","index":"","fulltext":""},{"type":"submitted","content":"npj Mental Health Research","date":"2025-04-28T10:15:53+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"npj-mental-health-research","isNatureJournal":false,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"npjmentalhealth","sideBox":"Learn more about [npj Mental Health Research](https://www.nature.com/npjmentalhealth/)","snPcode":"44184","submissionUrl":"https://mts-npjmentalhealth.nature.com/cgi-bin/main.p...","title":"npj Mental Health Research","twitterHandle":"@npjmentalhealth\n","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"npj","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"0eed6cca-eb61-4575-970c-e6e965d286d5","owner":[],"postedDate":"May 5th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":47868544,"name":"Health sciences/Health care"},{"id":47868545,"name":"Scientific community and society/Social sciences"},{"id":47868546,"name":"Scientific community and society/Social sciences/Psychology"}],"tags":[],"updatedAt":"2026-02-20T15:24:12+00:00","versionOfRecord":[],"versionCreatedAt":"2025-05-05 11:11:07","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6546781","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6546781","identity":"rs-6546781","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}