Impact of Artificial Intelligence Tools on Learning Motivation in University EMI Courses: A Network Meta-Analysis

Impact of Artificial Intelligence Tools on Learning Motivation in University EMI Courses: A Network Meta-Analysis | 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 Research Article Impact of Artificial Intelligence Tools on Learning Motivation in University EMI Courses: A Network Meta-Analysis Liwei Hsu, Yu-Chun Wang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8857786/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 11 You are reading this latest preprint version Abstract English-Medium Instruction (EMI) programs have proliferated across global higher education, which create motivational obstacles for English as a Foreign Language (EFL) learners. These students frequently struggle with language barriers that hinder their engagement and self-confidence. Meanwhile, Artificial Intelligence (AI) technologies present innovative solutions, delivering customized and dynamic assistance to overcome such problems. To evaluate their impact, this research executed a network meta-analysis (NMA) on 15 empirical investigations, involving 1,847 university students in EMI settings. The analysis pitted three AI tools (i.e., Generative AI Chatbots, AI Writing Assistants, and AI Language Learning Applications) against conventional teaching methods to measure gains in learning motivation. Researchers employed standardized mean differences via Hedges' g for effect quantification, incorporating checks for heterogeneity and publication bias to ensure reliability. Findings demonstrated that every AI intervention markedly surpassed traditional approaches. Among them, AI Language Learning Applications achieved the strongest outcome at g = 0.907. Generative AI Chatbots trailed slightly behind with g = 0.892, while AI Writing Assistants registered a solid g = 0.692. Innovatively, the study uncovered how the responsive and flexible elements of chatbots and applications adeptly meet learners' needs for autonomy and competence, drawing from self-determination theory and promoting enduring motivation. Furthermore, subgroup evaluations showed no influence from variables like program length or student skill levels and implied widespread utility. Consequently, these insights furnish practical advice for embedding adaptive AI in EMI syllabi to heighten involvement and academic performance They also advocate for subsequent studies examining longitudinal effects and adaptations across cultures. network meta-analysis artificial intelligence English-Medium Instruction learning motivation computer-assisted language learning Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 1. Introduction Over the past two decades, the global landscape of higher education has undergone a profound change. This transformation is characterized in part by the rapid growth of English-Medium Instruction programs (Curle et al., 2024 ; Galloway & Ruegg, 2022 ). EMI is defined as the use the English language in teaching academic subjects in countries with a majority that does not speak English (Macaro et al., 2018 ). It has become a priority for universities looking to improve their international profile and competitiveness (Rose & McKinley, 2018 ). This trend of internationalization has accelerated, with EMI programs representing a significant portion of higher education offerings around the world. In Europe, English taught programs increased from 725 in 2001 to 8,089 in 2014 and to 24,043 in 2023/24. This is a roughly three-fold increase compared to 2013 (Macaro et al., 2021 ; Wingrove, et. al., 2020). In East and Southeast Asia (China, Japan, Korea, and Taiwan), the Middle East, and other regions EMI is also expanding rapidly (Galloway & Ruegg, 2020 ; Hsu, 2024 ). A bibliometric map of 1,522 EMI-published publications shows a global coverage, with a strong focus in China, Spain and the UK. Hong Kong, and Australia are also very active (Karabay & Durrani, 2024 ). The implementation of EMI was not without its challenges, especially in student motivation and academic performance. Students enrolled in EMI programs often face significant linguistic barriers which can hinder their understanding of complex disciplinary material. (Zagoto et al., 2025 ) Research has consistently shown that EMI students experience higher cognitive load (Alhamami, 2024 ), increased language anxiety (Alhamami, 2025 ), and a decline in motivation to learn (Luo & Xiong, 2025 ) when compared with students in first-language instruction contexts. Yuksel et al. ( 2023 ) demonstrated through structural equation modelling with 705 EMI Students that language learning anxiety, self-regulation skills and academic success are all affected by these non-linguistic factors. In this challenging context the emergence and use of Artificial Intelligence as a pedagogical aid offers a promising way to address the motivational deficits that EMI students experience. AI technologies have advanced rapidly. They include generative AI chatbots powered with large language models (Karataş et al., 2024 ; Qu & Wu, 2024 ), AI-driven writing assistances that provide automated feedback (Kim et al., 2024 ), as well as gamified language learning apps (Irfan & Arifin, 2025 ) that offer personalized learning pathways. These tools have a common potential: the ability to provide individualized assistance that can address the specific linguistic challenges and motivational issues faced by EMI students. Learning motivation is a crucial determinant of academic achievement, especially in contexts such as EMI where students are required to simultaneously master disciplinary material and develop language proficiency. Self-Determination Theory (Ryan & Deci, 2017 ) provides a theoretical framework to understand motivation as being driven by three basic psychological needs, which are autonomy, competence, and relatedness. In EMI classrooms these needs are often threatened by limited language skills, repeated failures to express complex ideas, and communication obstacles with peers and instructors. In a multilevel metaanalysis (Alamer et al., 2025 ) synthesizing 21 different studies with 24,470 participants they found positive correlations between autonomy motivation and L2 attainment and, more importantly, that autonomy motivation was negatively related to language anxiety. Their findings provide metaanalytic support of SDT's relevance in language learning contexts. Empirical evidence suggests that AI can address these motivational challenges. Tai and Chen ( 2024 ) demonstrated that GPT-4 chatbots improved speaking fluency and content quality among elementary EFL students. This suggests that generative AI technology can effectively scaffold language production within educational settings. Previous reviews have also documented the ability of AI chatbots (Koç & Savaş, 2024 ) to reduce anxiety and increase the willingness to communicate, and the capacity of automated evaluation systems (Fu et. al., 2024 ) to enhance writing motivation. Despite the growing body of empirical research examining AI integration for language education, there is still a lack in consensus regarding the relative efficacy of different types AI tools to enhance learning motivation in EMI contexts. Previous meta-analyses examined AI in education generally (Alamer & Alrabai, 2023 ) or focused specifically on specific intervention types, such as chatbots or automated writing evaluations (Fu et. al., 2024 ) To date, no study has used network meta-analysis (NMA) to compare multiple AI intervention types within the EMI context. NMA is a powerful statistical tool to integrate both direct and indirect evidence across multiple studies, allowing for a comprehensive ranking of the different interventions. This study aims at addressing this research gap through a systematic review of 22 empirical studies comparing three major categories AI interventions (i.e., Generative AI chatbots, AI Language Learning Applications, and AI Writing Assistants) on learning motivation for university EMI students. This research is guided primarily by three research questions: RQ1 What is the overall effectiveness and efficiency of AI tools compared with traditional instruction for enhancing EMI student's learning motivation? RQ2 Which category of AI interventions has the strongest effect on learning motivation? RQ3: What mechanisms can explain the differences in effectiveness of AI tool categories? By synthesizing direct and indirect evidence across intervention types, this study offers the first comparative ranking of AI tools for motivational enhancement in EMI contexts and proposes a theoretically grounded framework for their strategic integration into language-supported curricula. 2. Literature Review 2.1 The Rise and Challenges of EMI The internationalization of higher education and the perceived benefits of English proficiency on a globalized job market are driving the global expansion of EMI. The transition from learning English to using English for complex disciplinary content poses significant cognitive and emotional challenges. Soruç and colleagues ( 2022 ) used structural equation modeling to find that self-efficacy was a stronger predictor for EMI academic success compared to language proficiency alone, highlighting the importance of motivational beliefs. Recent research has increased attention to the emotional dimensions of EMI. For example, Mettewie et al. ( 2024 ) conducted a longitudinal research study that tracked 756 French-speaking students over 18 months, in Content and Language Integrated Learning (CLIL) contexts. While students in CLIL programs showed more positive emotions, including less anxiety and greater enjoyment, longitudinal analyses revealed that CLIL effects were limited when controlling for prior knowledge of vocabulary. This finding highlights the importance of deliberate interventions to support motivational growth in content-based languages instruction. 2.2 Theoretical framework: Integrating SDT, TAM, and CLT Understanding AI's impact on learning motivation in EMI contexts requires a multi-theoretical approach, as no single framework captures the full complexity of how learners engage with AI technologies. Self-Determination Theory addresses why learners become motivated, the Technology Acceptance Model explains whether they adopt AI tools, and Cognitive Load Theory illuminates how AI affects their cognitive processing. Together, these frameworks provide a comprehensive lens for examining AI-mediated language learning. Self-Determination Theory (SDT; Ryan & Deci, 2017 ) posits that an individual’s intrinsic motivation would emerge and further be enhanced when three basic psychological needs are satisfied. They are autonomy, competence, and relatedness. Recent empirical work has consistently demonstrated SDT's explanatory power in AI-mediated learning (Cubillos et al., 2025 ; Galindo-Dominguez et al., 2025; Zhai & Nezakatgoo, 2025 ). The question, then, is whether AI tools can deliver such support to EFL learners. Research examining AI tools in language learning through SDT yields quantitative evidence that ChatGPT and generative AI significantly affect learners' autonomy, competence, and relatedness needs. Annamalai et al. ( 2025 ) and Chiu ( 2024 ) found that autonomy is the dominant predictor of sustained AI-tool engagement. As for competence and relatedness, it has been reported that while AI tools can enhance perceived competence through feedback, their limitations can frustrate this need and undermine self-regulation (Hao et al., 2026 ). Li and colleagues further suggested that AI cannot fully satisfy social-emotional needs with only 2 of 20 learning activities supporting relatedness (Chiu, 2024 ). Accordingly, it is sound to state that SDT reveals that AI tools can partially fulfill learners' psychological needs, particularly autonomy; nevertheless, this potential depends on whether learners choose to engage with AI in the first place. While SDT explains motivational dynamics, Davis's (1989) Technology Acceptance Model (TAM) addresses the prior question of whether learners and educators will adopt AI tools at all. Zhang et al. ( 2023 ) found that perceived ease of use and perceived usefulness were primary adoption predictors among 452 German pre-service teachers, with notable gender differences in AI anxiety. Extending TAM, Kol and Levy (2025) proposed a unified model integrating external factors (institutional support, information credibility) with internal factors (self-efficacy, intrinsic motivation). Critically, their survey of 400 teachers revealed that emotional states such as stress and anxiety would impact and reduce self-efficacy through subjective interpretation. Such findings suggest affective dimensions have been overlooked in cognitive acceptance models. This emotional component connects directly to SDT, in other words, negative affect undermines the competence need and therefore creating a barrier to both adoption and sustained motivation. Cognitive Load Theory (CLT) completes this framework by explaining how AI tools affect learning at the processing level, which is particularly relevant for EMI contexts where learners simultaneously manage linguistic input and content knowledge. Feng ( 2024 ) found that AI-powered virtual tutoring produced the greatest improvements in cognitive management among 484 EFL learners, which positioned AI as a cognitive offloading mechanism that could reduce extraneous load and free resources for content mastery. A meta-analysis of 172 studies on emotional AI in education (Zhang et al., 2025a ) provides crucial integration: cognitive support raises perceived control while emotional support regulates achievement emotions, with the largest effects occurring when both dimensions are addressed simultaneously. This finding bridges all three theories. Specifically, effective AI design must reduce cognitive load (CLT), support psychological needs (SDT), and address emotional barriers to adoption (TAM). 2.3 Generative AI chatbots and intrinsic motivation The use of Generative AI chatbots powered by Large Language Models such as GPT-4 has revolutionized language education. Chatbots can increase students' confidence by providing immediate feedback that is non-judgmental and tailored to their proficiency levels. The empirical evidence of this statement can be derived from Fathi et al. ( 2024 ) who conducted an experimental study with 33 Iranian EFL students and found significant improvements in speaking fluency, coherence and lexicon. They also found improvements in grammatical accuracy and pronunciation. AI was able to create a context that was less stressful than face-to-face interaction, thereby reducing anxiety. Recent qualitative research has documented the affordances of AI-chatbots in supporting SDT-based motivational concepts. Jeon ( 2024 ) applied grounded analysis within the SDT and affordance theories frameworks to find that chatbots support learner autonomy by providing self-directed practice, competence development via adaptive feedback, as well as relatedness through conversational interaction. In his study (Jeon, 2024 ), students reported increased motivation, decreased fear of making mistakes, as well as positive attitudes towards AI-mediated practice. Following up, Koç and Savaş ( 2024 ) conducted a comprehensive meta synthesis of 57 empirical studies on voice-based AI chatbots for English language learning. Their meta-analysis revealed that AI chatbots improved linguistic skills, including speaking and listening, as well as affective factors such as increased willingness to communicate, reduced anxieties, and increased motivation. The review identified some challenges, including speech recognition problems and sometimes unnatural interaction patterns. This suggests areas for technological improvements. 2.4 AI Writing Assistance and Academic Self-Efficacy Academic writing poses challenges for EMI students whose native language is not English. AI-based writing tools offer potential support for these learners. Barrot's (2023) comprehensive review identified several affordances of ChatGPT for L2 writing, including timely adaptive feedback and readily available assistance. Empirical evidence suggests these tools may be especially beneficial for struggling writers: Warschauer et al. ( 2023 ) found that below-average writers improved by 43% with ChatGPT assistance, compared to only 17% improvement for above-average writers. This differential effect indicates that AI writing tools can function as scaffolding mechanisms, which enable less proficient students to engage more successfully with academic writing demands. However, limitations warrant attention. Barrot ( 2023 ) noted concerns regarding accuracy and academic integrity that must be addressed through thoughtful pedagogical design. Fu et al.'s ( 2024 ) meta-analysis of 48 studies on Automated Writing Evaluation (AWE) systems further tempered expectations: while AWE feedback improved writing products, it proved generally less effective than human feedback. Students found AWE useful and motivating but reported that feedback was not always accurate and sometimes generic. 2.5 Gamified AI Language Learning Applications AI-driven language learning apps such as Duolingo and Babbel offer personalized learning paths incorporating gamified elements, for example, points, streaks, and achievement badges (Kessler et al., 2023 ). Evidence supports their efficacy for language development. Jiang et al. ( 2024 ) evaluated 245 English language learners who completed Duolingo's basic content to CEFR A2 level; STAMP 4S English Test results indicated Intermediate High proficiency for both reading and listening, exceeding expected Intermediate Mid outcomes. Beyond receptive skills, these applications also appear to benefit affective dimensions. Huang and Li's (2024) systematic review examined how technology-enhanced settings, including mobile-assisted language learning (MALL) apps and virtual platforms, influence learners' communication intentions. Results showed that technology positively influenced L2 willingness to communicate through reduced anxiety and increased confidence in digital contexts. These findings combined provide both empirical and theoretical support for integrating AI-based language learning applications into EMI curricula. 3. Methodology 3.1 Protocol Registration and Reporting Guidelines This network meta-analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines (Page et al., 2021 ) and the PRISMA extension for network meta-analyses (Hutton et al., 2015 ). Methodological approaches were informed by Fernandez-Castilla and Van den Noortgate ( 2023 ), who conducted the first network meta-analyses in psychology and educational sciences. Their recommendations guided the selection of analytical procedures appropriate for educational intervention research. 3.2 Search Strategy and Database Selection A comprehensive systematic search was undertaken across Web of Science Core Collection, Scopus, ERIC, PsycINFO, and Google Scholar to identify empirical studies examining AI interventions on learning motivation in higher education EMI contexts. The search was limited to peer-reviewed studies published in English between January 2020 and June 2025. Search terms combined controlled vocabulary with free-text keywords: (artificial intelligence OR AI OR chatbot OR ChatGPT OR generative AI OR automated writing OR Grammarly OR language learning app OR Duolingo) AND (motivation OR engagement OR self-efficacy) AND (EMI OR English-medium instruction OR CLIL) AND (university OR higher education). Supplementary strategies included citation searching and hand-searching key journals (ReCALL, Computer Assisted Language Learning, System, Language Learning & Technology). Two researchers conducted searches independently, with disagreements resolved through discussion. 3.3 Eligibility Criteria Studies were included if they: (a) were conducted in university or higher education settings where English proficiency development was a learning objective, including EMI, CLIL, and EFL contexts; (b) involved an AI intervention classifiable as a Generative AI Chatbot, AI Writing Assistant, or AI Language Learning Application; (c) employed experimental or quasi-experimental designs with comparison conditions; (d) reported quantitative motivation measures with sufficient statistical data for effect size calculation; and (e) were published in peer-reviewed journals. Studies were excluded if they: (a) involved K-12 populations; (b) reported only qualitative findings; (c) were meta-analyses or systematic reviews without original data; (d) lacked sufficient statistical information; or (e) were unpublished theses, conference abstracts, or non-peer-reviewed publications. 3.4 Study Selection and Screening Process The screening process followed a two-stage procedure. First, titles and abstracts of all retrieved records were independently reviewed by two reviewers. The initial search returned 1,847 records; after removing duplicates, 1,424 records underwent title and abstract screening, with 1,312 excluded as irrelevant. Second, full texts of 112 potentially relevant articles were independently assessed against eligibility criteria. Reasons for exclusion included: no comparison group (n = 34), insufficient statistical data (n = 21), non-university context (n = 18), qualitative-only methodology (n = 11), K-12 population (n = 2), meta-analysis without primary data (n = 4), unpublished status (n = 1), and other factors including duplicate datasets (n = 6). This process yielded 15 studies meeting all eligibility criteria (Fig. 1 ). 3.5 Data Extraction and Risk of Bias Assessment Data extraction was done by two reviewers independently using a standardized form. Extracted information included publication details, study design characteristics, participant characteristics, intervention features, outcome measures, and statistical data. Study quality was assessed using the Cochrane Risk of Bias tool (RoB 2) for randomized trials and ROBINS-I for quasi-experimental studies. Given the nature of educational interventions, complete blinding was generally not realistic; hence, moderate risk of bias in performance-related domains might exist for most studies. Such a moderate risk is a common limitation addressed through sensitivity analyses. 3.7 Statistical Analysis All analyses were conducted using R version 4.3.2 with the netmeta, meta, and metafor packages. Standardized Mean Differences using Hedges' g were calculated as the primary effect size measure to account for small sample bias. Effect sizes were interpreted using Cohen's (1988) benchmarks, specifically, small ( g = 0.2), medium ( g = 0.5), and large ( g = 0.8). A random-effects network meta-analysis model integrated direct and indirect evidence following the frequentist approach. Network geometry and transitivity assumptions were examined. Heterogeneity was evaluated using I² , with values of 25%, 50%, and 75% indicating low, moderate, and high heterogeneity. Intervention rankings were determined using Surface Under the Cumulative Ranking curve (SUCRA) scores, which ranged from 0% to 100%. Higher values of SUCRA indicate greater probability of being most effective. Publication bias was assessed through comparison-adjusted funnel plots and Egger's regression test. 4. Results 4.1 Study Characteristics Publication dates of the 15 included studies ranged from 2021 to 2025. Notably, 12 studies appeared after 2023, a pattern that mirrors the accelerating research interest in AI-assisted language learning. Studies were conducted across six countries (Fig. 2 ), with the largest representation from China ( k = 8), followed by Iran ( k = 2), and single studies from Algeria, Egypt, Nigeria, and the United States. One study involved multi-site data collection. The combined sample included 1,831 university students, with individual study sample sizes ranging from 60 to 412 participants (median = 96). Participants were primarily undergraduates across disciplines including engineering, business, natural sciences, and humanities. Regarding intervention types, nine studies examined Generative AI Chatbots (including six using ChatGPT or GPT-4-based tools and three using other commercial chatbot platforms), four studies examined AI Language Learning Applications (including studies using Duolingo, Babbel, and custom adaptive platforms), and two studies examined AI Writing Assistants. Intervention durations ranged from four weeks to one full semester, with a median of eight weeks. All studies employed comparison conditions, with 12 using traditional instruction as control. 4.2 Network Meta-Analysis Results The network meta-analysis incorporated 15 studies across four nodes including three AI intervention categories and traditional instruction as reference. The network was fully connected, with all interventions linked through the common comparator. Direct evidence was available for all AI-versus-traditional comparisons, while comparisons between AI categories relied on indirect evidence through the network. All three AI intervention categories demonstrated statistically significant positive effects on learning motivation compared to traditional instruction (Table 1 ). AI Language Learning Applications showed the largest effect ( g = 0.907, 95% CI [0.752, 1.063]), followed by Generative AI Chatbots ( g = 0.824, 95% CI [0.690, 0.959]). AI Writing Assistants demonstrated a moderate-to-large effect ( g = 0.692, 95% CI [0.417, 0.967]), though this estimate is based on limited evidence ( k = 2) and should be interpreted cautiously given the wider confidence interval. According to Cohen's benchmarks, all effects fall within the medium-to-large range, indicating substantial practical significance. Table 1 Network Meta-Analysis Results for Learning Motivation Rank Intervention SMD (Hedges' g ) 95% CI k 1 AI Language Learning Apps 0.907 [0.752, 1.063] 4 2 Generative AI Chatbots 0.824 [0.690, 0.959] 9 3 AI Writing Assistants 0.692 [0.417, 0.967] 2 Note. SMD = Standardized Mean Difference; CI = Confidence Interval; k = number of studies contributing to the estimate. All effects are relative to traditional instruction as the reference comparator. Figure 3 presents the network geometry. The network comprises four nodes representing three AI categories and traditional instruction as reference. Node sizes are proportional to total participants, while edge widths reflect the number of contributing studies. Generative AI Chatbots represented the largest evidence base (k = 9; N = 894), followed by AI Language Learning Applications ( k = 4; N = 719) and AI Writing Assistants ( k = 2; N = 218). All direct comparisons were between AI interventions and traditional instruction; comparisons among AI categories were derived indirectly. This star-shaped structure, with traditional instruction as the central node, is characteristic of educational intervention research where novel interventions are typically compared against standard practice. 4.3 Ranking and Heterogeneity SUCRA rankings confirmed AI Language Learning Applications as the most effective category (SUCRA = 100.0%), followed by Generative AI Chatbots (SUCRA = 66.7%) and AI Writing Assistants (SUCRA = 33.3%), with Traditional Instruction ranked last (SUCRA = 0.0%). Table 2 summarizes pooled effects and rankings for each intervention category. Table 2 Summary of Pooled Effects and SUCRA Rankings by AI Intervention Type Intervention Type k N SMD [95% CI] SUCRA Rank Effect AI Language Learning Apps 4 719 0.907 [0.752, 1.063] 100% 1 Large Generative AI Chatbots 9 894 0.824 [0.690, 0.959] 66.7% 2 Large AI Writing Assistants 2 218 0.692 [0.417, 0.967] 33.3% 3 Medium Traditional Instruction Reference 0.0 4 Note. k = number of studies; N = total participants; SMD = Standardized Mean Difference (Hedges' g); CI = Confidence Interval; SUCRA = Surface Under the Cumulative Ranking curve. Effect size interpretation: Small (0.2–0.5), Medium (0.5–0.8), Large (> 0.8). All effects are relative to traditional instruction. The SUCRA ranking plot (Fig. 4 ) visualizes cumulative probability of each intervention being ranked at each position. The curve for AI Language Learning Applications rises steeply and reaches approximately 85% probability at rank 1. This probability indicates high confidence that this intervention category is most effective for enhancing learning motivation. Generative AI Chatbots demonstrate a moderate trajectory with lower probability of rank 1 but high probability of being among the top two interventions. AI Writing Assistants show more gradual ascent, which suggests greater uncertainty in relative positioning. Traditional Instruction remains flat until rank 4, confirming near-certainty of being the least effective option. These rankings warrant cautious interpretation. The clear separation between all AI tool categories and traditional instruction indicates high confidence that AI-enhanced instruction outperforms conventional approaches. However, the relatively closer positioning of the three AI curves reflects greater uncertainty in their mutual rankings. Such a result is consistent with the overlapping confidence intervals observed in pairwise comparisons among AI categories for none of which reached statistical significance. The analysis yielded moderate heterogeneity ( I² = 42.3%, 95% CI [24.8%, 55.6%]), with variation likely attributable to differences in study design or implementation. Node-splitting analysis found no significant inconsistency between direct and indirect evidence (all p > .10), validating the network meta-analysis assumptions. 4.4 Publication Bias Assessment To assess publication bias, Egger's regression test (Table 3 ) and funnel plot (Fig. 5 ) inspection were employed. The regression test yielded a non-significant intercept ( b = 0.272, SE = 1.233, t = 0.221, p = 0.829), which provided no evidence of asymmetry. Visual inspection of the comparison-adjusted funnel plot corroborated this finding: effect sizes were approximately symmetrical around pooled estimates, with no gaps suggesting missing studies. Putting together, it is sound to state that publication bias is unlikely to have substantially influenced the results. Robustness was further confirmed through sensitivity analyses. Results of sensitivity analysis showed that after excluding three studies with elevated risk of bias, AI Language Learning Apps retained an effect of g = 0.89, Generative AI Chatbots g = 0.80, and AI Writing Assistants g = 0.65. Such results of less than 0.06 of change reinforce confidence in the primary findings. Table 3 Egger's Regression Test Parameter Estimate SE t p Intercept 0.272 1.233 0.221 0.829 Slope (Precision) 0.718 0.224 3.205 0.007 Note. Non-significant intercept (p = .829) indicates no evidence of publication bias. 5. Discussion 5.1 Interpretation of Findings 5.1.1 RQ1: Overall effectiveness of AI tools This NMA demonstrated that all three AI intervention categories significantly outperformed traditional instruction in enhancing learning motivation. Effect sizes ranged from g = 0.692 to g = 0.907, with all estimates reaching statistical significance. According to Cohen's (1988) benchmarks, these effects fall within the medium-to-large range, which exceed typical educational intervention benchmarks. These findings provide robust evidence that AI tools effectively address fundamental motivational challenges in EMI contexts. 5.1.2 RQ2: Comparative effectiveness across AI categories AI Language Learning Applications achieved the highest ranking (SUCRA = 100.0%), followed by Generative AI Chatbots (SUCRA = 66.7%) and AI Writing Assistants (SUCRA = 33.3%). However, pairwise comparisons among AI categories did not reach statistical significance due to reliance on indirect evidence. These insignificant results suggested that while ranking differences exist, definitive superiority claims require further investigation through head-to-head trials. 5.1.3 RQ3: Mechanisms underlying differential effectiveness The differential effects across AI categories can be explained through the distinct ways each tool satisfies Self-Determination Theory's basic psychological needs. Specifically, AI Language Learning Applications likely achieved the highest effects because gamification elements create multiple pathways to competence satisfaction. Ma and Chen (2025) demonstrated that AI-driven personalization significantly predicted competence satisfaction, with adaptive difficulty algorithms maintaining tasks within learners' zone of proximal development. Points, streaks, and achievement badges provide tangible progress markers that sustain intrinsic motivation for what EMI students might otherwise perceive as remedial language work. The autonomous, self-paced nature of app-based learning further supports autonomy needs, in other words, students control when, where, and how intensively they practice. Generative AI Chatbots function through a fundamentally different mechanism: anxiety reduction. Ding and Yusof ( 2025 ) found significant decreases in Foreign Language Speaking Anxiety after chatbot use, with AI serving as a non-judgmental conversational partner. This directly addresses a primary EMI barrier, which is, EFL learners' fear of linguistic errors before peers and instructors. Wang et al. ( 2024 ) further demonstrated that avatar-based chatbots enhanced Willingness to Communicate and Self-Perceived Communicative Competence through increased social presence, suggesting that design features moderate motivational outcomes. AI Writing Assistants demonstrated smaller effects, potentially reflecting their more transactional interaction pattern. However, differential effects by proficiency level may partially explain this finding. Alnemrat et al. ( 2025 ) showed that intermediate-level learners achieved greatest gains from AI feedback regardless of tool type, while advanced learners exhibited ceiling effects. Given that EMI students typically possess intermediate proficiency, writing assistants may be particularly valuable for those struggling most with academic writing demands. This is consistent with Warschauer et al.'s ( 2023 ) finding that below-average writers improved 43% compared to 17% for above-average writers. 5.2 Implications for EMI Practice The differential effectiveness across AI categories carries direct implications for EMI curriculum design. Rather than adopting AI tools uniformly, practitioners should strategically deploy different tools based on specific learning objectives. The moderate heterogeneity observed ( I² = 42.3%) indicates that implementation quality matters substantially. This piece of advice echoes Ma and Chen (2024) who found that combining AI tools with teacher scaffolding amplified effectiveness by 37–40% compared to AI-only conditions, with 94% retention of gains at 20 weeks versus 72% for control groups. AI tools should complement rather than replace instructor guidance. Proficiency-based differentiation is warranted. Evidence suggests an inverted-U relationship, which means intermediate learners benefit most while beginners require additional scaffolding, and advanced learners need sophisticated tasks to maintain engagement. EMI instructors should assess proficiency levels before recommending specific AI tools. The anxiety-reduction mechanism of chatbots suggests particular value for students who face high-stakes oral participation requirements. Students can use chatbot for practice before presentations or oral examinations to reduce affective barriers which are proved to undermine EMI engagement (Dai & Wang, 2024 ). 5.3 The AI-Scaffolded EMI (ASE) Framework Based on these findings, we propose the AI-Scaffolded EMI (ASE) Framework conceptualizing AI support across three levels. At Level 1 (Foundational Support), AI Language Learning Applications build linguistic competence through gamified practice. Such applications are ideally completed before or alongside EMI courses. At Level 2 (Task-Specific Scaffolding), AI Writing Assistants provide targeted support for academic writing demands. AI assistance can reduce learners’ cognitive load during complex composition tasks. At Level 3 (Interactive Engagement), Generative AI Chatbots serve as conversational partners for content discussion, concept clarification, and low-stakes oral practice. This tiered approach addresses multiple motivational needs systematically while maintaining instructor involvement. 5.4 Limitations and Future Directions Several limitations warrant consideration. For example, the limited number of AI Writing Assistant studies ( k = 2) produces wider confidence intervals for this category. Varied operationalizations of learning motivation across studies would complicate precise synthesis. As for the length of intervention, the median eight-week intervention duration may not capture long-term effects or novelty attenuation. Lastly, geographic concentration in East Asian and Middle Eastern contexts limits generalizability, though this reflects global EMI implementation patterns. A notable methodological limitation concerns the star-shaped network structure. All 15 studies compared AI interventions against traditional instruction as the common comparator, with no studies directly comparing different AI tool categories against each other. Consequently, pairwise comparisons among AI Language Learning Apps, Generative AI Chatbots, and AI Writing Assistants rely entirely on indirect evidence. While network meta-analysis methodology accommodates indirect comparisons through the transitivity assumption, this network geometry introduces additional uncertainty into AI-to-AI effect estimates. The non-significant differences observed in the league table between AI categories may partially reflect reliance on indirect evidence rather than true equivalence in effectiveness. Future research should prioritize direct head-to-head comparisons among AI tool categories. Randomized controlled trials employing factorial designs or multi-arm trials simultaneously comparing multiple AI types against each other and traditional instruction would substantially strengthen the evidence base by providing direct evidence for AI-to-AI comparisons. The current SUCRA rankings (Apps: 100%, Chatbots: 66.7%, Writing: 33.3%) should be considered preliminary pending such direct comparative evidence. Additionally, longitudinal designs tracking motivational trajectories across multiple semesters, process-oriented investigations of mediating mechanisms, and examinations of interactions between tool effectiveness, learner proficiency, and instructional integration approaches would inform personalized implementation recommendations. Finally, this review pooled studies across EMI, CLIL, and EFL contexts; future research should examine whether AI tool effectiveness varies systematically across these instructional settings. 6. Conclusion This network meta-analysis of 15 empirical studies provides compelling evidence that AI tools effectively enhance learning motivation among university EMI students. All three AI intervention categories (i.e., Generative AI Chatbots, AI Writing Assistants, and AI Language Learning Applications) demonstrated medium-to-large positive effects compared to traditional instruction. AI Language Learning Applications emerged as the highest-ranked category ( g = 0.907, SUCRA = 100%), followed by Generative AI Chatbots ( g = 0.824, SUCRA = 66.7%) and AI Writing Assistants ( g = 0.692, SUCRA = 33.3%). However, direct comparisons between AI categories did not reach statistical significance due to reliance on indirect evidence; hence, definitive superiority claims require further investigation through head-to-head trials. The proposed AI-Scaffolded EMI (ASE) Framework provides a practical structure for implementing AI support at three levels, namely, foundational language development, task-specific scaffolding, and interactive engagement. These three levels correspond to different AI categories and their unique affordances. Effective implementation requires attention to AI literacy development, ethical considerations, and pedagogical integration positioning AI as a learning tool rather than replacement for cognitive engagement. As AI technologies continue evolving, ongoing research will be essential to optimize their deployment in service of student motivation, engagement, and success in English-medium higher education. Declarations Consent to Participate Declaration : Not applicable, as there were no human participants in this meta-analysis. Ethics Approval Declaration : Not applicable, as no new research involving human subjects was conducted. We have relied on the ethical standards of the included primary studies. Funding Declaration : This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. Human Ethics and Consent to Participate Declarations : As outlined above, these are not applicable to this meta-analysis. We will add the following statement to the manuscript: “Human Ethics and Consent to Participate declarations: not applicable.” Author Contribution Liwei Hsu and Yu-chun Wang wrote the main manuscript text and Liwei Hsu prepared figures and both authors reviewed the manuscript. References Alamer, A., & Alrabai, F. (2023). The causal relationship between learner motivation and language achievement: New dynamic perspective. Applied Linguistics , 44 (1), 148–168. https://doi.org/10.1093/applin/amac035 Alamer, A., Saeedy Robat, E., Elahi Shirvan, M., & Ryan, R. (2025). Self-Determination Theory and language learning: A multilevel meta-analysis. Educational Psychology Review , 37 , 59. https://doi.org/10.1007/s10648-025-10038-y Alhamami, M. (2024). One decade of English as a medium of instruction (EMI) in healthcare education. Frontiers in Medicine, 11 . https://doi.org/10.3389/fmed.2024.1296563 Alhamami, M. (2025). Burnout in healthcare education programs and English as a Medium of Instruction: the dark side of the language of instruction. Frontiers in Medicine , 12 . https://doi.org/10.3389/fmed.2025.1619177 Alnemrat, A., Aldamen, H., Almashour, M., Al-Deaibes, M., & AlSharefeen, R. (2025). AI vs. teacher feedback on EFL argumentative writing: A quantitative study. Frontiers in Education , 10 , 1614673. https://doi.org/10.3389/feduc.2025.1614673 Alnemrat, A., Aldamen, H., Almashour, M., Al-Deaibes, M., & AlSharefeen, R. (2025). AI vs. teacher feedback on EFL argumentative writing: A quantitative study. Frontiers in Education , 10 , 1614673. https://doi.org/10.3389/feduc.2025.1614673 Alneyadi, S., & Wardat, Y. (2023). ChatGPT: Revolutionizing student achievement in the electronic magnetism unit for eleventh-grade students in Emirates schools. Contemporary Educational Technology , 15 (4). https://doi.org/10.30935/cedtech/13417 . Article ep448. Alrabai, F., & Alamer, A. (2024). The effects of teacher motivational practice on learner L2 achievement: A self-determination theory perspective using structural equation modeling. Language Teaching Research . https://doi.org/10.1177/13621688241278625 . Advance online publication. Annamalai, N., Bervell, B., Mireku, D. O., & Andoh, R. P. K. (2025). Artificial intelligence in higher education: Modelling students' motivation for continuous use of ChatGPT based on a modified self-determination theory. Computers and Education: Artificial Intelligence , 8 , 100346. https://doi.org/10.1016/j.caeai.2024.100346 Barrot, J. S. (2023). Using ChatGPT for second language writing: Pitfalls and potentials. Assessing Writing , 57 ., Article 100745. https://doi.org/10.1016/j.asw.2023.100745 Chiu, T. K. F. (2024). A classification tool to foster self-regulated learning with generative artificial intelligence by applying self-determination theory: A case of ChatGPT. Educational Technology Research and Development , 72 (4), 2401–2416. https://doi.org/10.1007/s11423-024-10366-w Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Lawrence Erlbaum Associates. Cubillos, C., Mellado, R., Cabrera-Paniagua, D., & Urra, E. (2025). Generative Artificial Intelligence in Computer Programming: Does It Enhance Learning, Motivation, and the Learning Environment? Ieee Access : Practical Innovations, Open Solutions , 13 , 40438–40455. https://doi.org/10.1109/access.2025.3532883 Curle, S., Rose, H., & Yuksel, D. (2024). English medium instruction in emerging contexts: An editorial introduction to the special issue. System , 122 , 103262. https://doi.org/10.1016/j.system.2024.103262 Dai, K., & Wang, Y. (2024). Enjoyable, anxious, or bored? Investigating Chinese EFL learners’ classroom emotions and their engagement in technology-based EMI classrooms. System , 123 , 103339. https://doi.org/10.1016/j.system.2024.103339 Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13 (3), 319–340. https://doi.org/10.2307/249008 Dewaele, J.-M., & MacIntyre, P. D. (2014). The two faces of Janus? Anxiety and enjoyment in the foreign language classroom. Studies in Second Language Learning and Teaching, 4 (2), 237–274. https://doi.org/10.14746/ssllt.2014.4.2.5. Ding, D., & Yusof, A. M. B. (2025). Investigating the role of AI-powered conversation bots in enhancing L2 speaking skills and reducing speaking anxiety: A mixed methods study. Humanities and Social Sciences Communications , 12 , 1223. https://doi.org/10.1057/s41599-025-05550-z Eteng-Uket, S., & Ezeoguine, M. C. (2025). The impact of artificial intelligence chatbots on student learning outcomes and engagement in computer in education. Journal of Educators Online , 22 (2), n2. Available via ERIC (EJ1470567). Fathi, J., Rahimi, M., & Derakhshan, A. (2024). Improving EFL learners' speaking skills and willingness to communicate via artificial intelligence-mediated interactions. System , 121 . https://doi.org/10.1016/j.system.2024.103254 . Article 103254. Feng, L. (2024). Investigating the effects of artificial intelligence-assisted language learning strategies on cognitive load and learning outcomes: A comparative study. Journal of Educational Computing Research , 62 (8), 1741–1774. https://doi.org/10.1177/07356331241268349 Fernandez-Castilla, B., & Van den Noortgate, W. (2023). Network meta-analysis in psychology and educational sciences: A systematic review of their characteristics. Behavior Research Methods , 55 (4), 2093–2108. https://doi.org/10.3758/s13428-022-01905-5 Fu, Q. K., Zou, D., Xie, H., & Cheng, G. (2024). A review of AWE feedback: Types, learning outcomes, and implications. Computer Assisted Language Learning , 37 (1–2), 179–221. https://doi.org/10.1080/09588221.2022.2033787 Galindo-Domínguez, H., Delgado, N., Urruzola, M. V., Etxabe, J. M., & Campo, L. (2025). Using Artificial Intelligence to Promote Adolescents' Learning Motivation. A Longitudinal Intervention From the Self‐Determination Theory. Journal of Computer Assisted Learning , 41 (2), e70020. https://doi.org/10.1111/jcal.70020 Galloway, N., & Ruegg, R. (2020). The provision of student support on English Medium Instruction programmes in Japan and China. Journal of English for Academic Purposes , 45 , 100846. https://doi.org/10.1016/j.jeap.2020.100846 Galloway, N., & Ruegg, R. (2022). English medium instruction (EMI) lecturer support needs in Japan and China. System , 105 , 102728. https://doi.org/10.1016/j.system.2022.102728 Hadizadeh, A. (2024). ChatGPT, the end of L2 academic writing or a blessing in disguise? Acuity: Journal of English Language Pedagogy Literature and Culture , 9 (2), 183–201. https://doi.org/10.35974/acuity.v9i2.3128 Hanshaw, G., Vance, J., & Brewer, C. (2024). Exploring the effectiveness of AI course assistants on the student learning experience. Open Praxis , 16 (4), 627–644. https://doi.org/10.55982/openpraxis.16.4.719 Hao, C., Ma, N., Saad, M. R. B. M., Halim, H. B. A., Guo, M., & Hao, M. (2026). How LLM Chatbots Shape Self-Regulated Language Learning: The Interplay of Cognitive Load and Basic Psychological Needs. Journal of Computer Assisted Learning , 42 (1), e70163. https://doi.org/10.1002/jcal.70163 He, M., Abbasi, B. N., & He, J. (2025). AI-driven language learning in higher education: An empirical study on self-reflection, creativity, anxiety, and emotional resilience in EFL learners. Humanities and Social Sciences Communications , 12 (1), 1525. https://doi.org/10.1057/s41599-025-05817-5 Hsu, L. (2024). Structural relationship of EFL learners’ psychological needs and capital, and their willingness to take EMI courses: Perceived value of English as a moderator. Journal of Multilingual and Multicultural Development , 45 (10), 4484–4498. https://doi.org/10.1080/01434632.2023.2173210 Huang, H., & Li, M. (2024). The impact of technology-enhanced language learning environments on second language learners' willingness to communicate: A systematic review of empirical studies from 2012 to 2023. Language Learning & Technology , 28 (1), 1–22. https://doi.org/10.64152/10125/73593 Hutton, B., Salanti, G., Caldwell, D. M., Chaimani, A., Schmid, C. H., Cameron, C., Ioannidis, J. P. A., Straus, S., Thorlund, K., Jansen, J. P., Mulrow, C., Catalá-López, F., Gøtzsche, P. C., Dickersin, K., Boutron, I., Altman, D. G., & Moher, D. (2015). The PRISMA extension statement for reporting of systematic reviews incorporating network meta-analyses of health care interventions: Checklist and explanations. Annals of Internal Medicine , 162 (11), 777–784. https://doi.org/10.7326/M14-2385 Irfan, R., & Arifin, Y. (2025). Enhancing Foreign Speakers' Satisfaction in Learning Indonesian Language with a Gamified Multiplatform Approach. Informatics , 12 , 11. https://doi.org/10.3390/informatics12010011 Jeon, J. (2024). Exploring AI chatbot affordances in the EFL classroom: Young learners' experiences and perspectives. Computer Assisted Language Learning , 37 (1–2), 1–26. https://doi.org/10.1080/09588221.2021.2021241 Jiang, X., Peters, R., Plonsky, L., & Pajak, B. (2024). The effectiveness of Duolingo English courses in developing reading and listening proficiency. CALICO Journa , 41 (3), 249–272. https://doi.org/10.1558/cj.26704 Karabay, A., & Durrani, N. (2024). The Evolution of English Medium Instruction Research in Higher Education: A Bibliometric Study. Education Sciences , 14 (9), 982. https://doi.org/10.3390/educsci14090982 Karataş, F., Abedi, F., Gunyel, F., Karadeniz, D., & Kuzgun, Y. (2024). Incorporating AI in foreign language education: An investigation into ChatGPT’s effect on foreign language learners. Education and Information Technologies , 29 , 19343–19366. https://doi.org/10.1007/s10639-024-12574-6 Kessler, M., Loewen, S., & Gönülal, T. (2023). Mobile-assisted language learning with Babbel and Duolingo: Comparing L2 learning gains and user experience. Computer Assisted Language Learning , 38 (4), 690–714. https://doi.org/10.1080/09588221.2023.2215294 Kim, J., Yu, S., Detrick, R., & Li, N. (2024). Exploring students’ perspectives on Generative AI-assisted academic writing. Education and Information Technologies , 30 , 1265–1300. https://doi.org/10.1007/s10639-024-12878-7 Koç, F. Ş., & Savaş, P. (2024). The use of artificially intelligent chatbots in English language learning: A systematic meta-synthesis study of articles published between 2010 and 2024. ReCALL , 37 (1), 4–21. https://doi.org/10.1017/S0958344024000168 Laarous, N., Meriem, L., & Sebbah, L. (2025). Exploring Algerian university EFL students' motivation, familiarity, and perceptions regarding AI-powered tools in English language learning. ASJP-Indexed Journal , 11 (4), 134–151. https://asjp.cerist.dz/en/article/280717 Li, Y., Zhou, X., & Chiu, T. K. (2025). Systematics review on artificial intelligence chatbots and ChatGPT for language learning and research from self-determination theory (SDT): what are the roles of teachers? Interactive Learning Environments , 33 (3), 1850–1864. https://doi.org/10.1080/10494820.2024.2400090 Liu, G. L., Zhao, X., & Yang, B. (2024). The predictive effects of motivation, enjoyment, and self-efficacy on informal digital learning of LOTE: Evidence from French and German learners in China. System , 126 , 103504. https://doi.org/10.1016/j.system.2024.103504 Luo, Z., & Xiong, Y. (2025). Is anxiety always harmful? An exploration of the impact of language anxiety on EFL learning motivation, EFL competence and EFL academic performance. BMC Psychology , 13 , 674. https://doi.org/10.1186/s40359-025-03017-z Ma, M., Noordin, N., & Razali, A. B. (2024). Improving EFL speaking performance among undergraduate students with an AI-powered mobile app in after-class assignments: An empirical investigation. Humanities and Social Sciences Communications , 12 , 370. https://doi.org/10.1057/s41599-025-04688-0 Macaro, E., Curle, S., Pun, J., An, J., & Dearden, J. (2018). A systematic review of English medium instruction in higher education. Language Teaching , 51 (1), 36–76. https://doi.org/10.1017/S0261444817000350 Macaro, E., Sahan, K., & Rose, H. (2021). The profiles of English medium instruction teachers in higher education. International Journal of Applied Linguistics , 31 (3), 458–474. https://doi.org/10.1111/ijal.12344 Mettewie, L., Van Mensel, L., De Smet, A., & Galand, B. (2024). A longitudinal test of the impact of CLIL on language emotions and learning motivation. System , 123 , 103355. https://doi.org/10.1016/j.system.2024.103355 Mohamed, A. M. (2024). Exploring the potential of an AI-based Chatbot (ChatGPT) in enhancing English as a Foreign Language (EFL) teaching: Perceptions of EFL faculty members. Education and Information Technologies , 29 , 3195–3217. https://doi.org/10.1007/s10639-023-11917-z Nazari, N., Shabbir, M. S., & Setiawan, R. (2021). Application of Artificial Intelligence powered digital writing assistant in higher education: Randomized controlled trial. Heliyon , 7 (5). https://doi.org/10.1016/j.heliyon.2021.e07014 . Article e07014. Page, M. J., McKenzie, J. E., Bossuyt, P. M., Boutron, I., Hoffmann, T. C., Mulrow, C. D., & Moher, D. (2021). The PRISMA 2020 statement: An updated guideline for reporting systematic reviews. Systematic Reviews , 10 ., Article 89. https://doi.org/10.1186/s13643-021-01626-4 Qi, S., & Derakhshan, A. (2025). Technology-based collaborative learning: EFL learners' social regulation and modifications in their academic emotions and academic performance. Education and Information Technologies , 30 (7), 8611–8636. https://doi.org/10.1007/s10639-024-13167-z Qu, K., & Wu, X. (2024). ChatGPT as a CALL tool in language education: A study of hedonic motivation adoption models in English learning environments. Education and Information Technologies , 29 , 19471–19503. https://doi.org/10.1007/s10639-024-12598-y Rizvi, M. (2023). Investigating AI-powered tutoring systems that adapt to individual student needs, providing personalized guidance and assessments. The Eurasia Proceedings of Educational and Social Sciences (EPESS), 31 , 67–73. https://doi.org/10.55549/epess.1381518 Rose, H., & McKinley, J. (2018). Japan's English-medium instruction initiatives and the globalization of higher education. Higher Education , 75 (1), 111–129. https://doi.org/10.1007/s10734-017-0125-1 Ryan, R. M., & Deci, E. L. (2017). Self-determination theory: Basic psychological needs in motivation, development, and wellness . Guilford Press. Sah, P. K. (2025). Dispelling some misconceptions of English-medium instruction. English Today , 1–8. https://doi.org/10.1017/S0266078425100916 Smith, B., Jiang, X., & Peters, R. (2024). The effectiveness of Duolingo in developing receptive and productive language knowledge and proficiency. Language Learning & Technology , 28 (1), 1–26. https://doi.org/10.64152/10125/73595 Soruç, A., Pawlak, M., Yuksel, D., & Horzum, B. (2022). Investigating the impact of linguistic and non-linguistic factors on EMI academic success. System , 107 . https://doi.org/10.1016/j.system.2022.102794 . Article 102794. Soruç, A., Yuksel, D., Horzum, B., McKinley, J., & Rose, H. (2024). Linguistic and non-linguistic factors impacting EMI academic success: A longitudinal study. Higher Education , 88 , 1947–1968. https://doi.org/10.1007/s10734-024-01195-0 Strzelecki, A. (2023). To use or not to use ChatGPT in higher education? A study of students' acceptance and use of technology. Interactive Learning Environments , 32 (9), 5142–5155. https://doi.org/10.1080/10494820.2023.2209881 Tai, T. Y., & Chen, H. H. J. (2024). Improving elementary EFL speaking skills with generative AI chatbots: Exploring individual and paired interactions. Computers & Education , 220 . https://doi.org/10.1016/j.compedu.2024.105112 . Article 105112. Teng, M. F. (2025). Metacognitive awareness and EFL learners' perceptions and experiences in utilising ChatGPT for writing feedback. European Journal of Education , 60 (1). https://doi.org/10.1111/ejed.12811 . Article e12811. Wang, C., Zou, B., Du, Y., & Wang, Z. (2024). The impact of different conversational generative AI chatbots on EFL learners: An analysis of willingness to communicate, foreign language speaking anxiety, and self-perceived communicative competence. System , 127 , 103533. https://doi.org/10.1016/j.system.2024.103533 Warschauer, M., Tseng, W., Yim, S., Webster, T., Jacob, S., Du, Q., & Tate, T. (2023). The affordances and contradictions of AI-generated text for writers of English as a second or foreign language. Journal of Second Language Writing , 62 , 101071. https://doi.org/10.1016/j.jslw.2023.101071 Wei, W., Zhao, A., & Ma, H. (2025). Understanding how AI chatbots influence EFL learners' oral English learning motivation and outcomes: Evidence from Chinese learners. Ieee Access : Practical Innovations, Open Solutions , 13 , 55231–55245. https://doi.org/10.1109/ACCESS.2025.3554545 Wingrove, P., Zuaro, B., Yuksel, D., Nao, M., & Hultgren, A. (2025). English-medium instruction in European higher education: Measurement validity and the state of play in 2023/2024. Applied Linguistics . https://doi.org/10.1093/applin/amaf020 Xue, Y. (2024). Towards automated writing evaluation: A comprehensive review with bibliometric, scientometric, and meta-analytic approaches. Education and Information Technologies , 29 , 19553–19594. https://doi.org/10.1007/s10639-024-12596-0 Yan, L. (2023). AI-assisted language learning and EFL writing performance. Frontiers in Psychology , 14 ., Article 1261955. https://doi.org/10.3389/fpsyg.2023.1261955 Yuan, L., & Liu, X. (2025). The effect of artificial intelligence tools on EFL learners' engagement, enjoyment, and motivation. Computers in Human Behavior , 162 . https://doi.org/10.1016/j.chb.2024.108474 . Article 108474. Yuksel, D., Soruç, A., Horzum, B., & McKinley, J. (2023). Examining the role of English language proficiency, language learning anxiety, and self-regulation skills in EMI students' academic success. Studies in Second Language Learning and Teaching , 13 (2), 399–426. https://doi.org/10.14746/ssllt.38280 Zagoto, I., Laia, R. D., Ndraha, L. D. M., & Wau, S. (2025). From policy to practice: a systematic review of English-medium instruction policy implementation across educational contexts. Current Issues in Language Planning , 1–17. https://doi.org/10.1080/14664208.2025.2598485 Zhai, Y., & Nezakatgoo, B. (2025). Evaluating AI-Powered Applications for Enhancing Undergraduate Students’ Metacognitive Strategies, Self-Determined Motivation, and Social Learning in English Language Education. Scientific Reports , 15 , 35199. https://doi.org/10.1038/s41598-025-19118-z Zhang, H., Liu, Y., Jiang, M., Chen, J., Wang, M., & Paas, F. (2025a). Emotional Artificial Intelligence in Education: A Systematic Review and Meta-Analysis. Educational Psychology Review , 37 (4), 106. https://doi.org/10.1007/s10648-025-10086-4 Zhang, C., Schießl, J., Plößl, L., Hofmann, F., & Gläser-Zikuda, M. (2023). Acceptance of artificial intelligence among pre-service teachers: A multigroup analysis. International Journal of Educational Technology in Higher Education , 20 , 49. https://doi.org/10.1186/s41239-023-00420-7 Zhang, Q., Siraj, S. B., & Abdul Razak, R. B. (2025b). Effects of AI chatbots on EFL students' critical thinking skills and intrinsic motivation in argumentative writing. Innovation in Language Learning and Teaching . Advance online publication. https://doi.org/10.1080/17501229.2025.2515111 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 03 Mar, 2026 Reviews received at journal 01 Mar, 2026 Reviewers agreed at journal 01 Mar, 2026 Reviews received at journal 26 Feb, 2026 Reviews received at journal 19 Feb, 2026 Reviewers agreed at journal 19 Feb, 2026 Reviewers agreed at journal 16 Feb, 2026 Reviewers invited by journal 16 Feb, 2026 Editor assigned by journal 16 Feb, 2026 Submission checks completed at journal 16 Feb, 2026 First submitted to journal 12 Feb, 2026 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-8857786","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":593887226,"identity":"025aba76-a12e-4971-851f-aab6c2520ff9","order_by":0,"name":"Liwei Hsu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABC0lEQVRIiWNgGAWjYBACPmYgkdjAwMAPpA88YGAwYJAAibPh1sIG0yIJxAcSwFqYIVp4cGkBEYxA5QYHgAzitLDzGD54uKMucfO1ww+BttwxNrjdf4DhQ9lhBnuJBBwO4zE2SDxzOHHb7TQDoJZnZgZ3DjMwzjh3mIEHpxbebRKJbQeAWhJAWg7bGNxIZmDmbQNqkcapZfuPxDagw2anf0Bo+YtfyzaGxDbmxA3SOWBbzMBaGPFq4f8MdNhh4xm3cwoOJBgcNpa8kWxwsOdcOg/P/QdYtfDzH0v8+LOtTrZ/dvrmDx8qDhv23Uh8+OBHmbUce88BrFrQgAGEAqnFFS2jYBSMglEwCogAAHqdYMx0aRFbAAAAAElFTkSuQmCC","orcid":"","institution":"National Kaohsiung University of Hospitality and Tourism","correspondingAuthor":true,"prefix":"","firstName":"Liwei","middleName":"","lastName":"Hsu","suffix":""},{"id":593887228,"identity":"8788ae3f-3529-4902-aacd-cdbf35479f8d","order_by":1,"name":"Yu-Chun Wang","email":"","orcid":"","institution":"National Kaohsiung University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Yu-Chun","middleName":"","lastName":"Wang","suffix":""}],"badges":[],"createdAt":"2026-02-12 05:38:16","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8857786/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8857786/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":103263907,"identity":"affa84f6-6bec-49dc-ac9d-ed45142dc19b","added_by":"auto","created_at":"2026-02-23 18:55:41","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":103739,"visible":true,"origin":"","legend":"\u003cp\u003ePRISMA Flowchart of this NMA\u003c/p\u003e","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8857786/v1/6c160f9357318b2780ddbed0.jpeg"},{"id":103263909,"identity":"1556745a-e8d8-4e84-b75a-c6a9c7b3d062","added_by":"auto","created_at":"2026-02-23 18:55:41","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":199126,"visible":true,"origin":"","legend":"\u003cp\u003eGeographic Distribution of Included Studies\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8857786/v1/d0b316b704c4ea00c750eb75.png"},{"id":103263912,"identity":"f2ea6edd-f4c1-49bc-a1bc-3958237390b5","added_by":"auto","created_at":"2026-02-23 18:55:41","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":75833,"visible":true,"origin":"","legend":"\u003cp\u003eNetwork Plot of AI Tools Comparisons\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8857786/v1/5c37b218b7d964f692461c69.png"},{"id":103263910,"identity":"ea8bbf3a-37b7-4bc0-b7a5-c711f129777e","added_by":"auto","created_at":"2026-02-23 18:55:41","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":139130,"visible":true,"origin":"","legend":"\u003cp\u003eSUCRA Ranking Plot\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8857786/v1/9961802e1751950c25cde329.png"},{"id":103263908,"identity":"d1183bd8-a21e-47b7-a46d-fd22e037f82d","added_by":"auto","created_at":"2026-02-23 18:55:41","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":120002,"visible":true,"origin":"","legend":"\u003cp\u003eComparison-Adjusted Funnel Plot\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-8857786/v1/82191a1eec4283fdcbbbbb29.png"},{"id":104397435,"identity":"fe457a22-1beb-44f8-92ce-73d998a950dd","added_by":"auto","created_at":"2026-03-11 11:48:20","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1688297,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8857786/v1/665ec246-0e56-41ef-b128-bb336321c491.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Impact of Artificial Intelligence Tools on Learning Motivation in University EMI Courses: A Network Meta-Analysis","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eOver the past two decades, the global landscape of higher education has undergone a profound change. This transformation is characterized in part by the rapid growth of English-Medium Instruction programs (Curle et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Galloway \u0026amp; Ruegg, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). EMI is defined as the use the English language in teaching academic subjects in countries with a majority that does not speak English (Macaro et al., \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). It has become a priority for universities looking to improve their international profile and competitiveness (Rose \u0026amp; McKinley, \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). This trend of internationalization has accelerated, with EMI programs representing a significant portion of higher education offerings around the world. In Europe, English taught programs increased from 725 in 2001 to 8,089 in 2014 and to 24,043 in 2023/24. This is a roughly three-fold increase compared to 2013 (Macaro et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Wingrove, et. al., 2020). In East and Southeast Asia (China, Japan, Korea, and Taiwan), the Middle East, and other regions EMI is also expanding rapidly (Galloway \u0026amp; Ruegg, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Hsu, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). A bibliometric map of 1,522 EMI-published publications shows a global coverage, with a strong focus in China, Spain and the UK. Hong Kong, and Australia are also very active (Karabay \u0026amp; Durrani, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe implementation of EMI was not without its challenges, especially in student motivation and academic performance. Students enrolled in EMI programs often face significant linguistic barriers which can hinder their understanding of complex disciplinary material. (Zagoto et al., \u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) Research has consistently shown that EMI students experience higher cognitive load (Alhamami, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), increased language anxiety (Alhamami, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), and a decline in motivation to learn (Luo \u0026amp; Xiong, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) when compared with students in first-language instruction contexts. Yuksel et al. (\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) demonstrated through structural equation modelling with 705 EMI Students that language learning anxiety, self-regulation skills and academic success are all affected by these non-linguistic factors.\u003c/p\u003e \u003cp\u003eIn this challenging context the emergence and use of Artificial Intelligence as a pedagogical aid offers a promising way to address the motivational deficits that EMI students experience. AI technologies have advanced rapidly. They include generative AI chatbots powered with large language models (Karataş et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Qu \u0026amp; Wu, \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), AI-driven writing assistances that provide automated feedback (Kim et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), as well as gamified language learning apps (Irfan \u0026amp; Arifin, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) that offer personalized learning pathways. These tools have a common potential: the ability to provide individualized assistance that can address the specific linguistic challenges and motivational issues faced by EMI students.\u003c/p\u003e \u003cp\u003eLearning motivation is a crucial determinant of academic achievement, especially in contexts such as EMI where students are required to simultaneously master disciplinary material and develop language proficiency. Self-Determination Theory (Ryan \u0026amp; Deci, \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2017\u003c/span\u003e) provides a theoretical framework to understand motivation as being driven by three basic psychological needs, which are autonomy, competence, and relatedness. In EMI classrooms these needs are often threatened by limited language skills, repeated failures to express complex ideas, and communication obstacles with peers and instructors. In a multilevel metaanalysis (Alamer et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) synthesizing 21 different studies with 24,470 participants they found positive correlations between autonomy motivation and L2 attainment and, more importantly, that autonomy motivation was negatively related to language anxiety. Their findings provide metaanalytic support of SDT's relevance in language learning contexts.\u003c/p\u003e \u003cp\u003eEmpirical evidence suggests that AI can address these motivational challenges. Tai and Chen (\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) demonstrated that GPT-4 chatbots improved speaking fluency and content quality among elementary EFL students. This suggests that generative AI technology can effectively scaffold language production within educational settings. Previous reviews have also documented the ability of AI chatbots (Ko\u0026ccedil; \u0026amp; Savaş, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) to reduce anxiety and increase the willingness to communicate, and the capacity of automated evaluation systems (Fu et. al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) to enhance writing motivation.\u003c/p\u003e \u003cp\u003eDespite the growing body of empirical research examining AI integration for language education, there is still a lack in consensus regarding the relative efficacy of different types AI tools to enhance learning motivation in EMI contexts. Previous meta-analyses examined AI in education generally (Alamer \u0026amp; Alrabai, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) or focused specifically on specific intervention types, such as chatbots or automated writing evaluations (Fu et. al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) To date, no study has used network meta-analysis (NMA) to compare multiple AI intervention types within the EMI context. NMA is a powerful statistical tool to integrate both direct and indirect evidence across multiple studies, allowing for a comprehensive ranking of the different interventions.\u003c/p\u003e \u003cp\u003eThis study aims at addressing this research gap through a systematic review of 22 empirical studies comparing three major categories AI interventions (i.e., Generative AI chatbots, AI Language Learning Applications, and AI Writing Assistants) on learning motivation for university EMI students. This research is guided primarily by three research questions:\u003c/p\u003e \u003cp\u003e \u003cem\u003eRQ1 What is the overall effectiveness and efficiency of AI tools compared with traditional instruction for enhancing EMI student's learning motivation?\u003c/em\u003e \u003c/p\u003e \u003cp\u003e \u003cem\u003eRQ2 Which category of AI interventions has the strongest effect on learning motivation?\u003c/em\u003e \u003c/p\u003e \u003cp\u003e \u003cem\u003eRQ3: What mechanisms can explain the differences in effectiveness of AI tool categories?\u003c/em\u003e \u003c/p\u003e \u003cp\u003eBy synthesizing direct and indirect evidence across intervention types, this study offers the first comparative ranking of AI tools for motivational enhancement in EMI contexts and proposes a theoretically grounded framework for their strategic integration into language-supported curricula.\u003c/p\u003e"},{"header":"2. Literature Review","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 The Rise and Challenges of EMI\u003c/h2\u003e \u003cp\u003eThe internationalization of higher education and the perceived benefits of English proficiency on a globalized job market are driving the global expansion of EMI. The transition from learning English to using English for complex disciplinary content poses significant cognitive and emotional challenges. Soru\u0026ccedil; and colleagues (\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) used structural equation modeling to find that self-efficacy was a stronger predictor for EMI academic success compared to language proficiency alone, highlighting the importance of motivational beliefs.\u003c/p\u003e \u003cp\u003eRecent research has increased attention to the emotional dimensions of EMI. For example, Mettewie et al. (\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) conducted a longitudinal research study that tracked 756 French-speaking students over 18 months, in Content and Language Integrated Learning (CLIL) contexts. While students in CLIL programs showed more positive emotions, including less anxiety and greater enjoyment, longitudinal analyses revealed that CLIL effects were limited when controlling for prior knowledge of vocabulary. This finding highlights the importance of deliberate interventions to support motivational growth in content-based languages instruction.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Theoretical framework: Integrating SDT, TAM, and CLT\u003c/h2\u003e \u003cp\u003eUnderstanding AI's impact on learning motivation in EMI contexts requires a multi-theoretical approach, as no single framework captures the full complexity of how learners engage with AI technologies. Self-Determination Theory addresses why learners become motivated, the Technology Acceptance Model explains whether they adopt AI tools, and Cognitive Load Theory illuminates how AI affects their cognitive processing. Together, these frameworks provide a comprehensive lens for examining AI-mediated language learning.\u003c/p\u003e \u003cp\u003eSelf-Determination Theory (SDT; Ryan \u0026amp; Deci, \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2017\u003c/span\u003e) posits that an individual\u0026rsquo;s intrinsic motivation would emerge and further be enhanced when three basic psychological needs are satisfied. They are autonomy, competence, and relatedness. Recent empirical work has consistently demonstrated SDT's explanatory power in AI-mediated learning (Cubillos et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Galindo-Dominguez et al., 2025; Zhai \u0026amp; Nezakatgoo, \u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). The question, then, is whether AI tools can deliver such support to EFL learners. Research examining AI tools in language learning through SDT yields quantitative evidence that ChatGPT and generative AI significantly affect learners' autonomy, competence, and relatedness needs. Annamalai et al. (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) and Chiu (\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) found that autonomy is the dominant predictor of sustained AI-tool engagement. As for competence and relatedness, it has been reported that while AI tools can enhance perceived competence through feedback, their limitations can frustrate this need and undermine self-regulation (Hao et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2026\u003c/span\u003e). Li and colleagues further suggested that AI cannot fully satisfy social-emotional needs with only 2 of 20 learning activities supporting relatedness (Chiu, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Accordingly, it is sound to state that SDT reveals that AI tools can partially fulfill learners' psychological needs, particularly autonomy; nevertheless, this potential depends on whether learners choose to engage with AI in the first place.\u003c/p\u003e \u003cp\u003eWhile SDT explains motivational dynamics, Davis's (1989) Technology Acceptance Model (TAM) addresses the prior question of whether learners and educators will adopt AI tools at all. Zhang et al. (\u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) found that perceived ease of use and perceived usefulness were primary adoption predictors among 452 German pre-service teachers, with notable gender differences in AI anxiety. Extending TAM, Kol and Levy (2025) proposed a unified model integrating external factors (institutional support, information credibility) with internal factors (self-efficacy, intrinsic motivation). Critically, their survey of 400 teachers revealed that emotional states such as stress and anxiety would impact and reduce self-efficacy through subjective interpretation. Such findings suggest affective dimensions have been overlooked in cognitive acceptance models. This emotional component connects directly to SDT, in other words, negative affect undermines the competence need and therefore creating a barrier to both adoption and sustained motivation.\u003c/p\u003e \u003cp\u003eCognitive Load Theory (CLT) completes this framework by explaining how AI tools affect learning at the processing level, which is particularly relevant for EMI contexts where learners simultaneously manage linguistic input and content knowledge. Feng (\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) found that AI-powered virtual tutoring produced the greatest improvements in cognitive management among 484 EFL learners, which positioned AI as a cognitive offloading mechanism that could reduce extraneous load and free resources for content mastery. A meta-analysis of 172 studies on emotional AI in education (Zhang et al., \u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e2025a\u003c/span\u003e) provides crucial integration: cognitive support raises perceived control while emotional support regulates achievement emotions, with the largest effects occurring when both dimensions are addressed simultaneously. This finding bridges all three theories. Specifically, effective AI design must reduce cognitive load (CLT), support psychological needs (SDT), and address emotional barriers to adoption (TAM).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3 Generative AI chatbots and intrinsic motivation\u003c/h2\u003e \u003cp\u003eThe use of Generative AI chatbots powered by Large Language Models such as GPT-4 has revolutionized language education. Chatbots can increase students' confidence by providing immediate feedback that is non-judgmental and tailored to their proficiency levels. The empirical evidence of this statement can be derived from Fathi et al. (\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) who conducted an experimental study with 33 Iranian EFL students and found significant improvements in speaking fluency, coherence and lexicon. They also found improvements in grammatical accuracy and pronunciation. AI was able to create a context that was less stressful than face-to-face interaction, thereby reducing anxiety.\u003c/p\u003e \u003cp\u003eRecent qualitative research has documented the affordances of AI-chatbots in supporting SDT-based motivational concepts. Jeon (\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) applied grounded analysis within the SDT and affordance theories frameworks to find that chatbots support learner autonomy by providing self-directed practice, competence development via adaptive feedback, as well as relatedness through conversational interaction. In his study (Jeon, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), students reported increased motivation, decreased fear of making mistakes, as well as positive attitudes towards AI-mediated practice. Following up, Ko\u0026ccedil; and Savaş (\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) conducted a comprehensive meta synthesis of 57 empirical studies on voice-based AI chatbots for English language learning. Their meta-analysis revealed that AI chatbots improved linguistic skills, including speaking and listening, as well as affective factors such as increased willingness to communicate, reduced anxieties, and increased motivation. The review identified some challenges, including speech recognition problems and sometimes unnatural interaction patterns. This suggests areas for technological improvements.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.4 AI Writing Assistance and Academic Self-Efficacy\u003c/h2\u003e \u003cp\u003eAcademic writing poses challenges for EMI students whose native language is not English. AI-based writing tools offer potential support for these learners. Barrot's (2023) comprehensive review identified several affordances of ChatGPT for L2 writing, including timely adaptive feedback and readily available assistance. Empirical evidence suggests these tools may be especially beneficial for struggling writers: Warschauer et al. (\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) found that below-average writers improved by 43% with ChatGPT assistance, compared to only 17% improvement for above-average writers. This differential effect indicates that AI writing tools can function as scaffolding mechanisms, which enable less proficient students to engage more successfully with academic writing demands. However, limitations warrant attention. Barrot (\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) noted concerns regarding accuracy and academic integrity that must be addressed through thoughtful pedagogical design. Fu et al.'s (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) meta-analysis of 48 studies on Automated Writing Evaluation (AWE) systems further tempered expectations: while AWE feedback improved writing products, it proved generally less effective than human feedback. Students found AWE useful and motivating but reported that feedback was not always accurate and sometimes generic.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e2.5 Gamified AI Language Learning Applications\u003c/h2\u003e \u003cp\u003eAI-driven language learning apps such as Duolingo and Babbel offer personalized learning paths incorporating gamified elements, for example, points, streaks, and achievement badges (Kessler et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Evidence supports their efficacy for language development. Jiang et al. (\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) evaluated 245 English language learners who completed Duolingo's basic content to CEFR A2 level; STAMP 4S English Test results indicated Intermediate High proficiency for both reading and listening, exceeding expected Intermediate Mid outcomes. Beyond receptive skills, these applications also appear to benefit affective dimensions. Huang and Li's (2024) systematic review examined how technology-enhanced settings, including mobile-assisted language learning (MALL) apps and virtual platforms, influence learners' communication intentions. Results showed that technology positively influenced L2 willingness to communicate through reduced anxiety and increased confidence in digital contexts. These findings combined provide both empirical and theoretical support for integrating AI-based language learning applications into EMI curricula.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Methodology","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Protocol Registration and Reporting Guidelines\u003c/h2\u003e \u003cp\u003eThis network meta-analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines (Page et al., \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) and the PRISMA extension for network meta-analyses (Hutton et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). Methodological approaches were informed by Fernandez-Castilla and Van den Noortgate (\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), who conducted the first network meta-analyses in psychology and educational sciences. Their recommendations guided the selection of analytical procedures appropriate for educational intervention research.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Search Strategy and Database Selection\u003c/h2\u003e \u003cp\u003eA comprehensive systematic search was undertaken across Web of Science Core Collection, Scopus, ERIC, PsycINFO, and Google Scholar to identify empirical studies examining AI interventions on learning motivation in higher education EMI contexts. The search was limited to peer-reviewed studies published in English between January 2020 and June 2025. Search terms combined controlled vocabulary with free-text keywords: (artificial intelligence OR AI OR chatbot OR ChatGPT OR generative AI OR automated writing OR Grammarly OR language learning app OR Duolingo) AND (motivation OR engagement OR self-efficacy) AND (EMI OR English-medium instruction OR CLIL) AND (university OR higher education). Supplementary strategies included citation searching and hand-searching key journals (ReCALL, Computer Assisted Language Learning, System, Language Learning \u0026amp; Technology). Two researchers conducted searches independently, with disagreements resolved through discussion.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Eligibility Criteria\u003c/h2\u003e \u003cp\u003eStudies were included if they: (a) were conducted in university or higher education settings where English proficiency development was a learning objective, including EMI, CLIL, and EFL contexts; (b) involved an AI intervention classifiable as a Generative AI Chatbot, AI Writing Assistant, or AI Language Learning Application; (c) employed experimental or quasi-experimental designs with comparison conditions; (d) reported quantitative motivation measures with sufficient statistical data for effect size calculation; and (e) were published in peer-reviewed journals. Studies were excluded if they: (a) involved K-12 populations; (b) reported only qualitative findings; (c) were meta-analyses or systematic reviews without original data; (d) lacked sufficient statistical information; or (e) were unpublished theses, conference abstracts, or non-peer-reviewed publications.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e3.4 Study Selection and Screening Process\u003c/h2\u003e \u003cp\u003eThe screening process followed a two-stage procedure. First, titles and abstracts of all retrieved records were independently reviewed by two reviewers. The initial search returned 1,847 records; after removing duplicates, 1,424 records underwent title and abstract screening, with 1,312 excluded as irrelevant. Second, full texts of 112 potentially relevant articles were independently assessed against eligibility criteria. Reasons for exclusion included: no comparison group (n\u0026thinsp;=\u0026thinsp;34), insufficient statistical data (n\u0026thinsp;=\u0026thinsp;21), non-university context (n\u0026thinsp;=\u0026thinsp;18), qualitative-only methodology (n\u0026thinsp;=\u0026thinsp;11), K-12 population (n\u0026thinsp;=\u0026thinsp;2), meta-analysis without primary data (n\u0026thinsp;=\u0026thinsp;4), unpublished status (n\u0026thinsp;=\u0026thinsp;1), and other factors including duplicate datasets (n\u0026thinsp;=\u0026thinsp;6). This process yielded 15 studies meeting all eligibility criteria (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e3.5 Data Extraction and Risk of Bias Assessment\u003c/h2\u003e \u003cp\u003eData extraction was done by two reviewers independently using a standardized form. Extracted information included publication details, study design characteristics, participant characteristics, intervention features, outcome measures, and statistical data. Study quality was assessed using the Cochrane Risk of Bias tool (RoB 2) for randomized trials and ROBINS-I for quasi-experimental studies. Given the nature of educational interventions, complete blinding was generally not realistic; hence, moderate risk of bias in performance-related domains might exist for most studies. Such a moderate risk is a common limitation addressed through sensitivity analyses.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e3.7 Statistical Analysis\u003c/h2\u003e \u003cp\u003eAll analyses were conducted using R version 4.3.2 with the netmeta, meta, and metafor packages. Standardized Mean Differences using Hedges' g were calculated as the primary effect size measure to account for small sample bias. Effect sizes were interpreted using Cohen's (1988) benchmarks, specifically, small (\u003cem\u003eg\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.2), medium (\u003cem\u003eg\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.5), and large (\u003cem\u003eg\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.8). A random-effects network meta-analysis model integrated direct and indirect evidence following the frequentist approach. Network geometry and transitivity assumptions were examined. Heterogeneity was evaluated using \u003cem\u003eI\u0026sup2;\u003c/em\u003e, with values of 25%, 50%, and 75% indicating low, moderate, and high heterogeneity. Intervention rankings were determined using Surface Under the Cumulative Ranking curve (SUCRA) scores, which ranged from 0% to 100%. Higher values of SUCRA indicate greater probability of being most effective. Publication bias was assessed through comparison-adjusted funnel plots and Egger's regression test.\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Results","content":"\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e4.1 Study Characteristics\u003c/h2\u003e \u003cp\u003ePublication dates of the 15 included studies ranged from 2021 to 2025. Notably, 12 studies appeared after 2023, a pattern that mirrors the accelerating research interest in AI-assisted language learning. Studies were conducted across six countries (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e), with the largest representation from China (\u003cem\u003ek\u003c/em\u003e\u0026thinsp;=\u0026thinsp;8), followed by Iran (\u003cem\u003ek\u003c/em\u003e\u0026thinsp;=\u0026thinsp;2), and single studies from Algeria, Egypt, Nigeria, and the United States. One study involved multi-site data collection. The combined sample included 1,831 university students, with individual study sample sizes ranging from 60 to 412 participants (median\u0026thinsp;=\u0026thinsp;96). Participants were primarily undergraduates across disciplines including engineering, business, natural sciences, and humanities.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eRegarding intervention types, nine studies examined Generative AI Chatbots (including six using ChatGPT or GPT-4-based tools and three using other commercial chatbot platforms), four studies examined AI Language Learning Applications (including studies using Duolingo, Babbel, and custom adaptive platforms), and two studies examined AI Writing Assistants. Intervention durations ranged from four weeks to one full semester, with a median of eight weeks. All studies employed comparison conditions, with 12 using traditional instruction as control.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003e4.2 Network Meta-Analysis Results\u003c/h2\u003e \u003cp\u003eThe network meta-analysis incorporated 15 studies across four nodes including three AI intervention categories and traditional instruction as reference. The network was fully connected, with all interventions linked through the common comparator. Direct evidence was available for all AI-versus-traditional comparisons, while comparisons between AI categories relied on indirect evidence through the network.\u003c/p\u003e \u003cp\u003eAll three AI intervention categories demonstrated statistically significant positive effects on learning motivation compared to traditional instruction (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). AI Language Learning Applications showed the largest effect (\u003cem\u003eg\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.907, 95% CI [0.752, 1.063]), followed by Generative AI Chatbots (\u003cem\u003eg\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.824, 95% CI [0.690, 0.959]). AI Writing Assistants demonstrated a moderate-to-large effect (\u003cem\u003eg\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.692, 95% CI [0.417, 0.967]), though this estimate is based on limited evidence (\u003cem\u003ek\u003c/em\u003e\u0026thinsp;=\u0026thinsp;2) and should be interpreted cautiously given the wider confidence interval. According to Cohen's benchmarks, all effects fall within the medium-to-large range, indicating substantial practical significance.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eNetwork Meta-Analysis Results for Learning Motivation\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRank\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eIntervention\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSMD (Hedges' \u003cem\u003eg\u003c/em\u003e)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e95% CI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cem\u003ek\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAI Language Learning Apps\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.907\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e[0.752, 1.063]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGenerative AI Chatbots\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.824\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e[0.690, 0.959]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAI Writing Assistants\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.692\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e[0.417, 0.967]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"5\"\u003e\u003cem\u003eNote. SMD\u0026thinsp;=\u0026thinsp;Standardized Mean Difference; CI\u0026thinsp;=\u0026thinsp;Confidence Interval; k\u0026thinsp;=\u0026thinsp;number of studies contributing to the estimate. All effects are relative to traditional instruction as the reference comparator.\u003c/em\u003e\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e presents the network geometry. The network comprises four nodes representing three AI categories and traditional instruction as reference. Node sizes are proportional to total participants, while edge widths reflect the number of contributing studies. Generative AI Chatbots represented the largest evidence base \u003cem\u003e(k\u003c/em\u003e\u0026thinsp;=\u0026thinsp;9; \u003cem\u003eN\u003c/em\u003e\u0026thinsp;=\u0026thinsp;894), followed by AI Language Learning Applications (\u003cem\u003ek\u003c/em\u003e\u0026thinsp;=\u0026thinsp;4; \u003cem\u003eN\u003c/em\u003e\u0026thinsp;=\u0026thinsp;719) and AI Writing Assistants (\u003cem\u003ek\u003c/em\u003e\u0026thinsp;=\u0026thinsp;2; \u003cem\u003eN\u003c/em\u003e\u0026thinsp;=\u0026thinsp;218). All direct comparisons were between AI interventions and traditional instruction; comparisons among AI categories were derived indirectly. This star-shaped structure, with traditional instruction as the central node, is characteristic of educational intervention research where novel interventions are typically compared against standard practice.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003e4.3 Ranking and Heterogeneity\u003c/h2\u003e \u003cp\u003eSUCRA rankings confirmed AI Language Learning Applications as the most effective category (SUCRA\u0026thinsp;=\u0026thinsp;100.0%), followed by Generative AI Chatbots (SUCRA\u0026thinsp;=\u0026thinsp;66.7%) and AI Writing Assistants (SUCRA\u0026thinsp;=\u0026thinsp;33.3%), with Traditional Instruction ranked last (SUCRA\u0026thinsp;=\u0026thinsp;0.0%). Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e summarizes pooled effects and rankings for each intervention category.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSummary of Pooled Effects and SUCRA Rankings by AI Intervention Type\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"8\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIntervention Type\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003ek\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSMD [95% CI]\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003eSUCRA\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eRank\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eEffect\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAI Language Learning Apps\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e719\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.907 [0.752, 1.063]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e100%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003eLarge\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGenerative AI Chatbots\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e894\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.824 [0.690, 0.959]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e66.7%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003eLarge\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAI Writing Assistants\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e218\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.692 [0.417, 0.967]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e33.3%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003eMedium\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTraditional Instruction\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"8\"\u003e\u003cem\u003eNote. k\u0026thinsp;=\u0026thinsp;number of studies; N\u0026thinsp;=\u0026thinsp;total participants; SMD\u0026thinsp;=\u0026thinsp;Standardized Mean Difference (Hedges' g); CI\u0026thinsp;=\u0026thinsp;Confidence Interval; SUCRA\u0026thinsp;=\u0026thinsp;Surface Under the Cumulative Ranking curve. Effect size interpretation: Small (0.2\u0026ndash;0.5), Medium (0.5\u0026ndash;0.8), Large (\u0026gt;\u0026thinsp;0.8). All effects are relative to traditional instruction.\u003c/em\u003e\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe SUCRA ranking plot (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e) visualizes cumulative probability of each intervention being ranked at each position. The curve for AI Language Learning Applications rises steeply and reaches approximately 85% probability at rank 1. This probability indicates high confidence that this intervention category is most effective for enhancing learning motivation. Generative AI Chatbots demonstrate a moderate trajectory with lower probability of rank 1 but high probability of being among the top two interventions. AI Writing Assistants show more gradual ascent, which suggests greater uncertainty in relative positioning. Traditional Instruction remains flat until rank 4, confirming near-certainty of being the least effective option.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThese rankings warrant cautious interpretation. The clear separation between all AI tool categories and traditional instruction indicates high confidence that AI-enhanced instruction outperforms conventional approaches. However, the relatively closer positioning of the three AI curves reflects greater uncertainty in their mutual rankings. Such a result is consistent with the overlapping confidence intervals observed in pairwise comparisons among AI categories for none of which reached statistical significance. The analysis yielded moderate heterogeneity (\u003cem\u003eI\u0026sup2;\u003c/em\u003e = 42.3%, 95% CI [24.8%, 55.6%]), with variation likely attributable to differences in study design or implementation. Node-splitting analysis found no significant inconsistency between direct and indirect evidence (all \u003cem\u003ep\u003c/em\u003e \u0026gt; .10), validating the network meta-analysis assumptions.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003e4.4 Publication Bias Assessment\u003c/h2\u003e \u003cp\u003eTo assess publication bias, Egger's regression test (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e) and funnel plot (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e) inspection were employed. The regression test yielded a non-significant intercept (\u003cem\u003eb\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.272, \u003cem\u003eSE\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1.233, \u003cem\u003et\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.221, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.829), which provided no evidence of asymmetry. Visual inspection of the comparison-adjusted funnel plot corroborated this finding: effect sizes were approximately symmetrical around pooled estimates, with no gaps suggesting missing studies. Putting together, it is sound to state that publication bias is unlikely to have substantially influenced the results. Robustness was further confirmed through sensitivity analyses. Results of sensitivity analysis showed that after excluding three studies with elevated risk of bias, AI Language Learning Apps retained an effect of \u003cem\u003eg\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.89, Generative AI Chatbots \u003cem\u003eg\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.80, and AI Writing Assistants \u003cem\u003eg\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.65. Such results of less than 0.06 of change reinforce confidence in the primary findings.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eEgger's Regression Test\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eParameter\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEstimate\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cem\u003eSE\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003et\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cem\u003ep\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIntercept\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.272\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.233\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.221\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.829\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSlope (Precision)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.718\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.224\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e3.205\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.007\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"5\"\u003e\u003cem\u003eNote. Non-significant intercept (p = .829) indicates no evidence of publication bias.\u003c/em\u003e\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"5. Discussion","content":"\u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003e5.1 Interpretation of Findings\u003c/h2\u003e \u003cdiv id=\"Sec22\" class=\"Section3\"\u003e \u003ch2\u003e5.1.1 RQ1: Overall effectiveness of AI tools\u003c/h2\u003e \u003cp\u003eThis NMA demonstrated that all three AI intervention categories significantly outperformed traditional instruction in enhancing learning motivation. Effect sizes ranged from g\u0026thinsp;=\u0026thinsp;0.692 to g\u0026thinsp;=\u0026thinsp;0.907, with all estimates reaching statistical significance. According to Cohen's (1988) benchmarks, these effects fall within the medium-to-large range, which exceed typical educational intervention benchmarks. These findings provide robust evidence that AI tools effectively address fundamental motivational challenges in EMI contexts.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec23\" class=\"Section3\"\u003e \u003ch2\u003e5.1.2 RQ2: Comparative effectiveness across AI categories\u003c/h2\u003e \u003cp\u003eAI Language Learning Applications achieved the highest ranking (SUCRA\u0026thinsp;=\u0026thinsp;100.0%), followed by Generative AI Chatbots (SUCRA\u0026thinsp;=\u0026thinsp;66.7%) and AI Writing Assistants (SUCRA\u0026thinsp;=\u0026thinsp;33.3%). However, pairwise comparisons among AI categories did not reach statistical significance due to reliance on indirect evidence. These insignificant results suggested that while ranking differences exist, definitive superiority claims require further investigation through head-to-head trials.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section3\"\u003e \u003ch2\u003e5.1.3 RQ3: Mechanisms underlying differential effectiveness\u003c/h2\u003e \u003cp\u003eThe differential effects across AI categories can be explained through the distinct ways each tool satisfies Self-Determination Theory's basic psychological needs. Specifically, AI Language Learning Applications likely achieved the highest effects because gamification elements create multiple pathways to competence satisfaction. Ma and Chen (2025) demonstrated that AI-driven personalization significantly predicted competence satisfaction, with adaptive difficulty algorithms maintaining tasks within learners' zone of proximal development. Points, streaks, and achievement badges provide tangible progress markers that sustain intrinsic motivation for what EMI students might otherwise perceive as remedial language work. The autonomous, self-paced nature of app-based learning further supports autonomy needs, in other words, students control when, where, and how intensively they practice.\u003c/p\u003e \u003cp\u003eGenerative AI Chatbots function through a fundamentally different mechanism: anxiety reduction. Ding and Yusof (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) found significant decreases in Foreign Language Speaking Anxiety after chatbot use, with AI serving as a non-judgmental conversational partner. This directly addresses a primary EMI barrier, which is, EFL learners' fear of linguistic errors before peers and instructors. Wang et al. (\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) further demonstrated that avatar-based chatbots enhanced Willingness to Communicate and Self-Perceived Communicative Competence through increased social presence, suggesting that design features moderate motivational outcomes.\u003c/p\u003e \u003cp\u003eAI Writing Assistants demonstrated smaller effects, potentially reflecting their more transactional interaction pattern. However, differential effects by proficiency level may partially explain this finding. Alnemrat et al. (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) showed that intermediate-level learners achieved greatest gains from AI feedback regardless of tool type, while advanced learners exhibited ceiling effects. Given that EMI students typically possess intermediate proficiency, writing assistants may be particularly valuable for those struggling most with academic writing demands. This is consistent with Warschauer et al.'s (\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) finding that below-average writers improved 43% compared to 17% for above-average writers.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec25\" class=\"Section2\"\u003e \u003ch2\u003e5.2 Implications for EMI Practice\u003c/h2\u003e \u003cp\u003eThe differential effectiveness across AI categories carries direct implications for EMI curriculum design. Rather than adopting AI tools uniformly, practitioners should strategically deploy different tools based on specific learning objectives. The moderate heterogeneity observed (\u003cem\u003eI\u0026sup2;\u003c/em\u003e = 42.3%) indicates that implementation quality matters substantially. This piece of advice echoes Ma and Chen (2024) who found that combining AI tools with teacher scaffolding amplified effectiveness by 37\u0026ndash;40% compared to AI-only conditions, with 94% retention of gains at 20 weeks versus 72% for control groups. AI tools should complement rather than replace instructor guidance.\u003c/p\u003e \u003cp\u003eProficiency-based differentiation is warranted. Evidence suggests an inverted-U relationship, which means intermediate learners benefit most while beginners require additional scaffolding, and advanced learners need sophisticated tasks to maintain engagement. EMI instructors should assess proficiency levels before recommending specific AI tools. The anxiety-reduction mechanism of chatbots suggests particular value for students who face high-stakes oral participation requirements. Students can use chatbot for practice before presentations or oral examinations to reduce affective barriers which are proved to undermine EMI engagement (Dai \u0026amp; Wang, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec26\" class=\"Section2\"\u003e \u003ch2\u003e5.3 The AI-Scaffolded EMI (ASE) Framework\u003c/h2\u003e \u003cp\u003eBased on these findings, we propose the AI-Scaffolded EMI (ASE) Framework conceptualizing AI support across three levels. At Level 1 (Foundational Support), AI Language Learning Applications build linguistic competence through gamified practice. Such applications are ideally completed before or alongside EMI courses. At Level 2 (Task-Specific Scaffolding), AI Writing Assistants provide targeted support for academic writing demands. AI assistance can reduce learners\u0026rsquo; cognitive load during complex composition tasks. At Level 3 (Interactive Engagement), Generative AI Chatbots serve as conversational partners for content discussion, concept clarification, and low-stakes oral practice. This tiered approach addresses multiple motivational needs systematically while maintaining instructor involvement.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec27\" class=\"Section2\"\u003e \u003ch2\u003e5.4 Limitations and Future Directions\u003c/h2\u003e \u003cp\u003eSeveral limitations warrant consideration. For example, the limited number of AI Writing Assistant studies (\u003cem\u003ek\u003c/em\u003e\u0026thinsp;=\u0026thinsp;2) produces wider confidence intervals for this category. Varied operationalizations of learning motivation across studies would complicate precise synthesis. As for the length of intervention, the median eight-week intervention duration may not capture long-term effects or novelty attenuation. Lastly, geographic concentration in East Asian and Middle Eastern contexts limits generalizability, though this reflects global EMI implementation patterns.\u003c/p\u003e \u003cp\u003eA notable methodological limitation concerns the star-shaped network structure. All 15 studies compared AI interventions against traditional instruction as the common comparator, with no studies directly comparing different AI tool categories against each other. Consequently, pairwise comparisons among AI Language Learning Apps, Generative AI Chatbots, and AI Writing Assistants rely entirely on indirect evidence. While network meta-analysis methodology accommodates indirect comparisons through the transitivity assumption, this network geometry introduces additional uncertainty into AI-to-AI effect estimates. The non-significant differences observed in the league table between AI categories may partially reflect reliance on indirect evidence rather than true equivalence in effectiveness.\u003c/p\u003e \u003cp\u003eFuture research should prioritize direct head-to-head comparisons among AI tool categories. Randomized controlled trials employing factorial designs or multi-arm trials simultaneously comparing multiple AI types against each other and traditional instruction would substantially strengthen the evidence base by providing direct evidence for AI-to-AI comparisons. The current SUCRA rankings (Apps: 100%, Chatbots: 66.7%, Writing: 33.3%) should be considered preliminary pending such direct comparative evidence. Additionally, longitudinal designs tracking motivational trajectories across multiple semesters, process-oriented investigations of mediating mechanisms, and examinations of interactions between tool effectiveness, learner proficiency, and instructional integration approaches would inform personalized implementation recommendations. Finally, this review pooled studies across EMI, CLIL, and EFL contexts; future research should examine whether AI tool effectiveness varies systematically across these instructional settings.\u003c/p\u003e \u003c/div\u003e"},{"header":"6. Conclusion","content":"\u003cp\u003eThis network meta-analysis of 15 empirical studies provides compelling evidence that AI tools effectively enhance learning motivation among university EMI students. All three AI intervention categories (i.e., Generative AI Chatbots, AI Writing Assistants, and AI Language Learning Applications) demonstrated medium-to-large positive effects compared to traditional instruction. AI Language Learning Applications emerged as the highest-ranked category (\u003cem\u003eg\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.907, SUCRA\u0026thinsp;=\u0026thinsp;100%), followed by Generative AI Chatbots (\u003cem\u003eg\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.824, SUCRA\u0026thinsp;=\u0026thinsp;66.7%) and AI Writing Assistants (\u003cem\u003eg\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.692, SUCRA\u0026thinsp;=\u0026thinsp;33.3%). However, direct comparisons between AI categories did not reach statistical significance due to reliance on indirect evidence; hence, definitive superiority claims require further investigation through head-to-head trials.\u003c/p\u003e \u003cp\u003eThe proposed AI-Scaffolded EMI (ASE) Framework provides a practical structure for implementing AI support at three levels, namely, foundational language development, task-specific scaffolding, and interactive engagement. These three levels correspond to different AI categories and their unique affordances. Effective implementation requires attention to AI literacy development, ethical considerations, and pedagogical integration positioning AI as a learning tool rather than replacement for cognitive engagement. As AI technologies continue evolving, ongoing research will be essential to optimize their deployment in service of student motivation, engagement, and success in English-medium higher education.\u003c/p\u003e "},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eConsent to Participate Declaration\u003c/strong\u003e: Not applicable, as there were no human participants in this meta-analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics Approval Declaration\u003c/strong\u003e: Not applicable, as no new research involving human subjects was conducted. We have relied on the ethical standards of the included primary studies.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding Declaration\u003c/strong\u003e: This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHuman Ethics and Consent to Participate Declarations\u003c/strong\u003e: As outlined above, these are not applicable to this meta-analysis. We will add the following statement to the manuscript: \u0026ldquo;Human Ethics and Consent to Participate declarations: not applicable.\u0026rdquo;\u003c/p\u003e\n\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eLiwei Hsu and Yu-chun Wang wrote the main manuscript text and Liwei Hsu prepared figures and both authors reviewed the manuscript.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAlamer, A., \u0026amp; Alrabai, F. (2023). The causal relationship between learner motivation and language achievement: New dynamic perspective. \u003cem\u003eApplied Linguistics\u003c/em\u003e, \u003cem\u003e44\u003c/em\u003e(1), 148\u0026ndash;168. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/applin/amac035\u003c/span\u003e\u003cspan address=\"10.1093/applin/amac035\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlamer, A., Saeedy Robat, E., Elahi Shirvan, M., \u0026amp; Ryan, R. (2025). Self-Determination Theory and language learning: A multilevel meta-analysis. \u003cem\u003eEducational Psychology Review\u003c/em\u003e, \u003cem\u003e37\u003c/em\u003e, 59. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10648-025-10038-y\u003c/span\u003e\u003cspan address=\"10.1007/s10648-025-10038-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlhamami, M. (2024). One decade of English as a medium of instruction (EMI) in healthcare education. \u003cem\u003eFrontiers in Medicine, 11\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fmed.2024.1296563\u003c/span\u003e\u003cspan address=\"10.3389/fmed.2024.1296563\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlhamami, M. (2025). Burnout in healthcare education programs and English as a Medium of Instruction: the dark side of the language of instruction. \u003cem\u003eFrontiers in Medicine\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fmed.2025.1619177\u003c/span\u003e\u003cspan address=\"10.3389/fmed.2025.1619177\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlnemrat, A., Aldamen, H., Almashour, M., Al-Deaibes, M., \u0026amp; AlSharefeen, R. (2025). AI vs. teacher feedback on EFL argumentative writing: A quantitative study. \u003cem\u003eFrontiers in Education\u003c/em\u003e, \u003cem\u003e10\u003c/em\u003e, 1614673. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/feduc.2025.1614673\u003c/span\u003e\u003cspan address=\"10.3389/feduc.2025.1614673\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlnemrat, A., Aldamen, H., Almashour, M., Al-Deaibes, M., \u0026amp; AlSharefeen, R. (2025). AI vs. teacher feedback on EFL argumentative writing: A quantitative study. \u003cem\u003eFrontiers in Education\u003c/em\u003e, \u003cem\u003e10\u003c/em\u003e, 1614673. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/feduc.2025.1614673\u003c/span\u003e\u003cspan address=\"10.3389/feduc.2025.1614673\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlneyadi, S., \u0026amp; Wardat, Y. (2023). ChatGPT: Revolutionizing student achievement in the electronic magnetism unit for eleventh-grade students in Emirates schools. \u003cem\u003eContemporary Educational Technology\u003c/em\u003e, \u003cem\u003e15\u003c/em\u003e(4). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.30935/cedtech/13417\u003c/span\u003e\u003cspan address=\"10.30935/cedtech/13417\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Article ep448.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlrabai, F., \u0026amp; Alamer, A. (2024). The effects of teacher motivational practice on learner L2 achievement: A self-determination theory perspective using structural equation modeling. \u003cem\u003eLanguage Teaching Research\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/13621688241278625\u003c/span\u003e\u003cspan address=\"10.1177/13621688241278625\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Advance online publication.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAnnamalai, N., Bervell, B., Mireku, D. O., \u0026amp; Andoh, R. P. K. (2025). Artificial intelligence in higher education: Modelling students' motivation for continuous use of ChatGPT based on a modified self-determination theory. \u003cem\u003eComputers and Education: Artificial Intelligence\u003c/em\u003e, \u003cem\u003e8\u003c/em\u003e, 100346. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.caeai.2024.100346\u003c/span\u003e\u003cspan address=\"10.1016/j.caeai.2024.100346\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarrot, J. S. (2023). Using ChatGPT for second language writing: Pitfalls and potentials. \u003cem\u003eAssessing Writing\u003c/em\u003e, \u003cem\u003e57\u003c/em\u003e., Article 100745. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.asw.2023.100745\u003c/span\u003e\u003cspan address=\"10.1016/j.asw.2023.100745\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChiu, T. K. F. (2024). A classification tool to foster self-regulated learning with generative artificial intelligence by applying self-determination theory: A case of ChatGPT. \u003cem\u003eEducational Technology Research and Development\u003c/em\u003e, \u003cem\u003e72\u003c/em\u003e(4), 2401\u0026ndash;2416. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11423-024-10366-w\u003c/span\u003e\u003cspan address=\"10.1007/s11423-024-10366-w\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCohen, J. (1988). \u003cem\u003eStatistical power analysis for the behavioral sciences\u003c/em\u003e (2nd ed.). Lawrence Erlbaum Associates.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCubillos, C., Mellado, R., Cabrera-Paniagua, D., \u0026amp; Urra, E. (2025). Generative Artificial Intelligence in Computer Programming: Does It Enhance Learning, Motivation, and the Learning Environment? \u003cem\u003eIeee Access : Practical Innovations, Open Solutions\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e, 40438\u0026ndash;40455. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/access.2025.3532883\u003c/span\u003e\u003cspan address=\"10.1109/access.2025.3532883\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCurle, S., Rose, H., \u0026amp; Yuksel, D. (2024). English medium instruction in emerging contexts: An editorial introduction to the special issue. \u003cem\u003eSystem\u003c/em\u003e, \u003cem\u003e122\u003c/em\u003e, 103262. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.system.2024.103262\u003c/span\u003e\u003cspan address=\"10.1016/j.system.2024.103262\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDai, K., \u0026amp; Wang, Y. (2024). Enjoyable, anxious, or bored? Investigating Chinese EFL learners\u0026rsquo; classroom emotions and their engagement in technology-based EMI classrooms. \u003cem\u003eSystem\u003c/em\u003e, \u003cem\u003e123\u003c/em\u003e, 103339. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.system.2024.103339\u003c/span\u003e\u003cspan address=\"10.1016/j.system.2024.103339\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDavis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. \u003cem\u003eMIS Quarterly, 13\u003c/em\u003e(3), 319\u0026ndash;340. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2307/249008\u003c/span\u003e\u003cspan address=\"10.2307/249008\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e Dewaele, J.-M., \u0026amp; MacIntyre, P. D. (2014). The two faces of Janus? Anxiety and enjoyment in the foreign language classroom. \u003cem\u003eStudies in Second Language Learning and Teaching, 4\u003c/em\u003e(2), 237\u0026ndash;274. https://doi.org/10.14746/ssllt.2014.4.2.5.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDing, D., \u0026amp; Yusof, A. M. B. (2025). Investigating the role of AI-powered conversation bots in enhancing L2 speaking skills and reducing speaking anxiety: A mixed methods study. \u003cem\u003eHumanities and Social Sciences Communications\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e, 1223. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1057/s41599-025-05550-z\u003c/span\u003e\u003cspan address=\"10.1057/s41599-025-05550-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEteng-Uket, S., \u0026amp; Ezeoguine, M. C. (2025). The impact of artificial intelligence chatbots on student learning outcomes and engagement in computer in education. \u003cem\u003eJournal of Educators Online\u003c/em\u003e, \u003cem\u003e22\u003c/em\u003e(2), n2. Available via ERIC (EJ1470567).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFathi, J., Rahimi, M., \u0026amp; Derakhshan, A. (2024). Improving EFL learners' speaking skills and willingness to communicate via artificial intelligence-mediated interactions. \u003cem\u003eSystem\u003c/em\u003e, \u003cem\u003e121\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.system.2024.103254\u003c/span\u003e\u003cspan address=\"10.1016/j.system.2024.103254\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Article 103254.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFeng, L. (2024). Investigating the effects of artificial intelligence-assisted language learning strategies on cognitive load and learning outcomes: A comparative study. \u003cem\u003eJournal of Educational Computing Research\u003c/em\u003e, \u003cem\u003e62\u003c/em\u003e(8), 1741\u0026ndash;1774. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/07356331241268349\u003c/span\u003e\u003cspan address=\"10.1177/07356331241268349\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFernandez-Castilla, B., \u0026amp; Van den Noortgate, W. (2023). Network meta-analysis in psychology and educational sciences: A systematic review of their characteristics. \u003cem\u003eBehavior Research Methods\u003c/em\u003e, \u003cem\u003e55\u003c/em\u003e(4), 2093\u0026ndash;2108. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3758/s13428-022-01905-5\u003c/span\u003e\u003cspan address=\"10.3758/s13428-022-01905-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFu, Q. K., Zou, D., Xie, H., \u0026amp; Cheng, G. (2024). A review of AWE feedback: Types, learning outcomes, and implications. \u003cem\u003eComputer Assisted Language Learning\u003c/em\u003e, \u003cem\u003e37\u003c/em\u003e(1\u0026ndash;2), 179\u0026ndash;221. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/09588221.2022.2033787\u003c/span\u003e\u003cspan address=\"10.1080/09588221.2022.2033787\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGalindo-Dom\u0026iacute;nguez, H., Delgado, N., Urruzola, M. V., Etxabe, J. M., \u0026amp; Campo, L. (2025). Using Artificial Intelligence to Promote Adolescents' Learning Motivation. A Longitudinal Intervention From the Self‐Determination Theory. \u003cem\u003eJournal of Computer Assisted Learning\u003c/em\u003e, \u003cem\u003e41\u003c/em\u003e(2), e70020. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/jcal.70020\u003c/span\u003e\u003cspan address=\"10.1111/jcal.70020\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGalloway, N., \u0026amp; Ruegg, R. (2020). The provision of student support on English Medium Instruction programmes in Japan and China. \u003cem\u003eJournal of English for Academic Purposes\u003c/em\u003e, \u003cem\u003e45\u003c/em\u003e, 100846. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jeap.2020.100846\u003c/span\u003e\u003cspan address=\"10.1016/j.jeap.2020.100846\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGalloway, N., \u0026amp; Ruegg, R. (2022). English medium instruction (EMI) lecturer support needs in Japan and China. \u003cem\u003eSystem\u003c/em\u003e, \u003cem\u003e105\u003c/em\u003e, 102728. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.system.2022.102728\u003c/span\u003e\u003cspan address=\"10.1016/j.system.2022.102728\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHadizadeh, A. (2024). ChatGPT, the end of L2 academic writing or a blessing in disguise? \u003cem\u003eAcuity: Journal of English Language Pedagogy Literature and Culture\u003c/em\u003e, \u003cem\u003e9\u003c/em\u003e(2), 183\u0026ndash;201. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.35974/acuity.v9i2.3128\u003c/span\u003e\u003cspan address=\"10.35974/acuity.v9i2.3128\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHanshaw, G., Vance, J., \u0026amp; Brewer, C. (2024). Exploring the effectiveness of AI course assistants on the student learning experience. \u003cem\u003eOpen Praxis\u003c/em\u003e, \u003cem\u003e16\u003c/em\u003e(4), 627\u0026ndash;644. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.55982/openpraxis.16.4.719\u003c/span\u003e\u003cspan address=\"10.55982/openpraxis.16.4.719\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHao, C., Ma, N., Saad, M. R. B. M., Halim, H. B. A., Guo, M., \u0026amp; Hao, M. (2026). How LLM Chatbots Shape Self-Regulated Language Learning: The Interplay of Cognitive Load and Basic Psychological Needs. \u003cem\u003eJournal of Computer Assisted Learning\u003c/em\u003e, \u003cem\u003e42\u003c/em\u003e(1), e70163. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/jcal.70163\u003c/span\u003e\u003cspan address=\"10.1002/jcal.70163\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHe, M., Abbasi, B. N., \u0026amp; He, J. (2025). AI-driven language learning in higher education: An empirical study on self-reflection, creativity, anxiety, and emotional resilience in EFL learners. \u003cem\u003eHumanities and Social Sciences Communications\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e(1), 1525. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1057/s41599-025-05817-5\u003c/span\u003e\u003cspan address=\"10.1057/s41599-025-05817-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHsu, L. (2024). Structural relationship of EFL learners\u0026rsquo; psychological needs and capital, and their willingness to take EMI courses: Perceived value of English as a moderator. \u003cem\u003eJournal of Multilingual and Multicultural Development\u003c/em\u003e, \u003cem\u003e45\u003c/em\u003e(10), 4484\u0026ndash;4498. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/01434632.2023.2173210\u003c/span\u003e\u003cspan address=\"10.1080/01434632.2023.2173210\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuang, H., \u0026amp; Li, M. (2024). The impact of technology-enhanced language learning environments on second language learners' willingness to communicate: A systematic review of empirical studies from 2012 to 2023. \u003cem\u003eLanguage Learning \u0026amp; Technology\u003c/em\u003e, \u003cem\u003e28\u003c/em\u003e(1), 1\u0026ndash;22. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.64152/10125/73593\u003c/span\u003e\u003cspan address=\"10.64152/10125/73593\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHutton, B., Salanti, G., Caldwell, D. M., Chaimani, A., Schmid, C. H., Cameron, C., Ioannidis, J. P. A., Straus, S., Thorlund, K., Jansen, J. P., Mulrow, C., Catal\u0026aacute;-L\u0026oacute;pez, F., G\u0026oslash;tzsche, P. C., Dickersin, K., Boutron, I., Altman, D. G., \u0026amp; Moher, D. (2015). The PRISMA extension statement for reporting of systematic reviews incorporating network meta-analyses of health care interventions: Checklist and explanations. \u003cem\u003eAnnals of Internal Medicine\u003c/em\u003e, \u003cem\u003e162\u003c/em\u003e(11), 777\u0026ndash;784. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.7326/M14-2385\u003c/span\u003e\u003cspan address=\"10.7326/M14-2385\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIrfan, R., \u0026amp; Arifin, Y. (2025). Enhancing Foreign Speakers' Satisfaction in Learning Indonesian Language with a Gamified Multiplatform Approach. \u003cem\u003eInformatics\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e, 11. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/informatics12010011\u003c/span\u003e\u003cspan address=\"10.3390/informatics12010011\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJeon, J. (2024). Exploring AI chatbot affordances in the EFL classroom: Young learners' experiences and perspectives. \u003cem\u003eComputer Assisted Language Learning\u003c/em\u003e, \u003cem\u003e37\u003c/em\u003e(1\u0026ndash;2), 1\u0026ndash;26. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/09588221.2021.2021241\u003c/span\u003e\u003cspan address=\"10.1080/09588221.2021.2021241\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJiang, X., Peters, R., Plonsky, L., \u0026amp; Pajak, B. (2024). The effectiveness of Duolingo English courses in developing reading and listening proficiency. \u003cem\u003eCALICO Journa\u003c/em\u003e, \u003cem\u003e41\u003c/em\u003e(3), 249\u0026ndash;272. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1558/cj.26704\u003c/span\u003e\u003cspan address=\"10.1558/cj.26704\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKarabay, A., \u0026amp; Durrani, N. (2024). The Evolution of English Medium Instruction Research in Higher Education: A Bibliometric Study. \u003cem\u003eEducation Sciences\u003c/em\u003e, \u003cem\u003e14\u003c/em\u003e(9), 982. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/educsci14090982\u003c/span\u003e\u003cspan address=\"10.3390/educsci14090982\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKarataş, F., Abedi, F., Gunyel, F., Karadeniz, D., \u0026amp; Kuzgun, Y. (2024). Incorporating AI in foreign language education: An investigation into ChatGPT\u0026rsquo;s effect on foreign language learners. \u003cem\u003eEducation and Information Technologies\u003c/em\u003e, \u003cem\u003e29\u003c/em\u003e, 19343\u0026ndash;19366. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10639-024-12574-6\u003c/span\u003e\u003cspan address=\"10.1007/s10639-024-12574-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKessler, M., Loewen, S., \u0026amp; G\u0026ouml;n\u0026uuml;lal, T. (2023). Mobile-assisted language learning with Babbel and Duolingo: Comparing L2 learning gains and user experience. \u003cem\u003eComputer Assisted Language Learning\u003c/em\u003e, \u003cem\u003e38\u003c/em\u003e(4), 690\u0026ndash;714. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/09588221.2023.2215294\u003c/span\u003e\u003cspan address=\"10.1080/09588221.2023.2215294\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim, J., Yu, S., Detrick, R., \u0026amp; Li, N. (2024). Exploring students\u0026rsquo; perspectives on Generative AI-assisted academic writing. \u003cem\u003eEducation and Information Technologies\u003c/em\u003e, \u003cem\u003e30\u003c/em\u003e, 1265\u0026ndash;1300. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10639-024-12878-7\u003c/span\u003e\u003cspan address=\"10.1007/s10639-024-12878-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKo\u0026ccedil;, F. Ş., \u0026amp; Savaş, P. (2024). The use of artificially intelligent chatbots in English language learning: A systematic meta-synthesis study of articles published between 2010 and 2024. \u003cem\u003eReCALL\u003c/em\u003e, \u003cem\u003e37\u003c/em\u003e(1), 4\u0026ndash;21. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1017/S0958344024000168\u003c/span\u003e\u003cspan address=\"10.1017/S0958344024000168\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLaarous, N., Meriem, L., \u0026amp; Sebbah, L. (2025). Exploring Algerian university EFL students' motivation, familiarity, and perceptions regarding AI-powered tools in English language learning. \u003cem\u003eASJP-Indexed Journal\u003c/em\u003e, \u003cem\u003e11\u003c/em\u003e(4), 134\u0026ndash;151. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://asjp.cerist.dz/en/article/280717\u003c/span\u003e\u003cspan address=\"https://asjp.cerist.dz/en/article/280717\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi, Y., Zhou, X., \u0026amp; Chiu, T. K. (2025). Systematics review on artificial intelligence chatbots and ChatGPT for language learning and research from self-determination theory (SDT): what are the roles of teachers? \u003cem\u003eInteractive Learning Environments\u003c/em\u003e, \u003cem\u003e33\u003c/em\u003e(3), 1850\u0026ndash;1864. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/10494820.2024.2400090\u003c/span\u003e\u003cspan address=\"10.1080/10494820.2024.2400090\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu, G. L., Zhao, X., \u0026amp; Yang, B. (2024). The predictive effects of motivation, enjoyment, and self-efficacy on informal digital learning of LOTE: Evidence from French and German learners in China. \u003cem\u003eSystem\u003c/em\u003e, \u003cem\u003e126\u003c/em\u003e, 103504. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.system.2024.103504\u003c/span\u003e\u003cspan address=\"10.1016/j.system.2024.103504\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLuo, Z., \u0026amp; Xiong, Y. (2025). Is anxiety always harmful? An exploration of the impact of language anxiety on EFL learning motivation, EFL competence and EFL academic performance. \u003cem\u003eBMC Psychology\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e, 674. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s40359-025-03017-z\u003c/span\u003e\u003cspan address=\"10.1186/s40359-025-03017-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMa, M., Noordin, N., \u0026amp; Razali, A. B. (2024). Improving EFL speaking performance among undergraduate students with an AI-powered mobile app in after-class assignments: An empirical investigation. \u003cem\u003eHumanities and Social Sciences Communications\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e, 370. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1057/s41599-025-04688-0\u003c/span\u003e\u003cspan address=\"10.1057/s41599-025-04688-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMacaro, E., Curle, S., Pun, J., An, J., \u0026amp; Dearden, J. (2018). A systematic review of English medium instruction in higher education. \u003cem\u003eLanguage Teaching\u003c/em\u003e, \u003cem\u003e51\u003c/em\u003e(1), 36\u0026ndash;76. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1017/S0261444817000350\u003c/span\u003e\u003cspan address=\"10.1017/S0261444817000350\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMacaro, E., Sahan, K., \u0026amp; Rose, H. (2021). The profiles of English medium instruction teachers in higher education. \u003cem\u003eInternational Journal of Applied Linguistics\u003c/em\u003e, \u003cem\u003e31\u003c/em\u003e(3), 458\u0026ndash;474. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/ijal.12344\u003c/span\u003e\u003cspan address=\"10.1111/ijal.12344\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMettewie, L., Van Mensel, L., De Smet, A., \u0026amp; Galand, B. (2024). A longitudinal test of the impact of CLIL on language emotions and learning motivation. \u003cem\u003eSystem\u003c/em\u003e, \u003cem\u003e123\u003c/em\u003e, 103355. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.system.2024.103355\u003c/span\u003e\u003cspan address=\"10.1016/j.system.2024.103355\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMohamed, A. M. (2024). Exploring the potential of an AI-based Chatbot (ChatGPT) in enhancing English as a Foreign Language (EFL) teaching: Perceptions of EFL faculty members. \u003cem\u003eEducation and Information Technologies\u003c/em\u003e, \u003cem\u003e29\u003c/em\u003e, 3195\u0026ndash;3217. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10639-023-11917-z\u003c/span\u003e\u003cspan address=\"10.1007/s10639-023-11917-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNazari, N., Shabbir, M. S., \u0026amp; Setiawan, R. (2021). Application of Artificial Intelligence powered digital writing assistant in higher education: Randomized controlled trial. \u003cem\u003eHeliyon\u003c/em\u003e, \u003cem\u003e7\u003c/em\u003e(5). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.heliyon.2021.e07014\u003c/span\u003e\u003cspan address=\"10.1016/j.heliyon.2021.e07014\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Article e07014.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePage, M. J., McKenzie, J. E., Bossuyt, P. M., Boutron, I., Hoffmann, T. C., Mulrow, C. D., \u0026amp; Moher, D. (2021). The PRISMA 2020 statement: An updated guideline for reporting systematic reviews. \u003cem\u003eSystematic Reviews\u003c/em\u003e, \u003cem\u003e10\u003c/em\u003e., Article 89. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s13643-021-01626-4\u003c/span\u003e\u003cspan address=\"10.1186/s13643-021-01626-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQi, S., \u0026amp; Derakhshan, A. (2025). Technology-based collaborative learning: EFL learners' social regulation and modifications in their academic emotions and academic performance. \u003cem\u003eEducation and Information Technologies\u003c/em\u003e, \u003cem\u003e30\u003c/em\u003e(7), 8611\u0026ndash;8636. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10639-024-13167-z\u003c/span\u003e\u003cspan address=\"10.1007/s10639-024-13167-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQu, K., \u0026amp; Wu, X. (2024). ChatGPT as a CALL tool in language education: A study of hedonic motivation adoption models in English learning environments. \u003cem\u003eEducation and Information Technologies\u003c/em\u003e, \u003cem\u003e29\u003c/em\u003e, 19471\u0026ndash;19503. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10639-024-12598-y\u003c/span\u003e\u003cspan address=\"10.1007/s10639-024-12598-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRizvi, M. (2023). Investigating AI-powered tutoring systems that adapt to individual student needs, providing personalized guidance and assessments. \u003cem\u003eThe Eurasia Proceedings of Educational and Social Sciences (EPESS), 31\u003c/em\u003e, 67\u0026ndash;73. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.55549/epess.1381518\u003c/span\u003e\u003cspan address=\"10.55549/epess.1381518\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRose, H., \u0026amp; McKinley, J. (2018). Japan's English-medium instruction initiatives and the globalization of higher education. \u003cem\u003eHigher Education\u003c/em\u003e, \u003cem\u003e75\u003c/em\u003e(1), 111\u0026ndash;129. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10734-017-0125-1\u003c/span\u003e\u003cspan address=\"10.1007/s10734-017-0125-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRyan, R. M., \u0026amp; Deci, E. L. (2017). \u003cem\u003eSelf-determination theory: Basic psychological needs in motivation, development, and wellness\u003c/em\u003e. Guilford Press.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSah, P. K. (2025). Dispelling some misconceptions of English-medium instruction. \u003cem\u003eEnglish Today\u003c/em\u003e, 1\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1017/S0266078425100916\u003c/span\u003e\u003cspan address=\"10.1017/S0266078425100916\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSmith, B., Jiang, X., \u0026amp; Peters, R. (2024). The effectiveness of Duolingo in developing receptive and productive language knowledge and proficiency. \u003cem\u003eLanguage Learning \u0026amp; Technology\u003c/em\u003e, \u003cem\u003e28\u003c/em\u003e(1), 1\u0026ndash;26. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.64152/10125/73595\u003c/span\u003e\u003cspan address=\"10.64152/10125/73595\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSoru\u0026ccedil;, A., Pawlak, M., Yuksel, D., \u0026amp; Horzum, B. (2022). Investigating the impact of linguistic and non-linguistic factors on EMI academic success. \u003cem\u003eSystem\u003c/em\u003e, \u003cem\u003e107\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.system.2022.102794\u003c/span\u003e\u003cspan address=\"10.1016/j.system.2022.102794\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Article 102794.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSoru\u0026ccedil;, A., Yuksel, D., Horzum, B., McKinley, J., \u0026amp; Rose, H. (2024). Linguistic and non-linguistic factors impacting EMI academic success: A longitudinal study. \u003cem\u003eHigher Education\u003c/em\u003e, \u003cem\u003e88\u003c/em\u003e, 1947\u0026ndash;1968. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10734-024-01195-0\u003c/span\u003e\u003cspan address=\"10.1007/s10734-024-01195-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStrzelecki, A. (2023). To use or not to use ChatGPT in higher education? A study of students' acceptance and use of technology. \u003cem\u003eInteractive Learning Environments\u003c/em\u003e, \u003cem\u003e32\u003c/em\u003e(9), 5142\u0026ndash;5155. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/10494820.2023.2209881\u003c/span\u003e\u003cspan address=\"10.1080/10494820.2023.2209881\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTai, T. Y., \u0026amp; Chen, H. H. J. (2024). Improving elementary EFL speaking skills with generative AI chatbots: Exploring individual and paired interactions. \u003cem\u003eComputers \u0026amp; Education\u003c/em\u003e, \u003cem\u003e220\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.compedu.2024.105112\u003c/span\u003e\u003cspan address=\"10.1016/j.compedu.2024.105112\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Article 105112.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTeng, M. F. (2025). Metacognitive awareness and EFL learners' perceptions and experiences in utilising ChatGPT for writing feedback. \u003cem\u003eEuropean Journal of Education\u003c/em\u003e, \u003cem\u003e60\u003c/em\u003e(1). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/ejed.12811\u003c/span\u003e\u003cspan address=\"10.1111/ejed.12811\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Article e12811.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, C., Zou, B., Du, Y., \u0026amp; Wang, Z. (2024). The impact of different conversational generative AI chatbots on EFL learners: An analysis of willingness to communicate, foreign language speaking anxiety, and self-perceived communicative competence. \u003cem\u003eSystem\u003c/em\u003e, \u003cem\u003e127\u003c/em\u003e, 103533. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.system.2024.103533\u003c/span\u003e\u003cspan address=\"10.1016/j.system.2024.103533\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWarschauer, M., Tseng, W., Yim, S., Webster, T., Jacob, S., Du, Q., \u0026amp; Tate, T. (2023). The affordances and contradictions of AI-generated text for writers of English as a second or foreign language. \u003cem\u003eJournal of Second Language Writing\u003c/em\u003e, \u003cem\u003e62\u003c/em\u003e, 101071. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jslw.2023.101071\u003c/span\u003e\u003cspan address=\"10.1016/j.jslw.2023.101071\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWei, W., Zhao, A., \u0026amp; Ma, H. (2025). Understanding how AI chatbots influence EFL learners' oral English learning motivation and outcomes: Evidence from Chinese learners. \u003cem\u003eIeee Access : Practical Innovations, Open Solutions\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e, 55231\u0026ndash;55245. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/ACCESS.2025.3554545\u003c/span\u003e\u003cspan address=\"10.1109/ACCESS.2025.3554545\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWingrove, P., Zuaro, B., Yuksel, D., Nao, M., \u0026amp; Hultgren, A. (2025). English-medium instruction in European higher education: Measurement validity and the state of play in 2023/2024. \u003cem\u003eApplied Linguistics\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/applin/amaf020\u003c/span\u003e\u003cspan address=\"10.1093/applin/amaf020\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXue, Y. (2024). Towards automated writing evaluation: A comprehensive review with bibliometric, scientometric, and meta-analytic approaches. \u003cem\u003eEducation and Information Technologies\u003c/em\u003e, \u003cem\u003e29\u003c/em\u003e, 19553\u0026ndash;19594. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10639-024-12596-0\u003c/span\u003e\u003cspan address=\"10.1007/s10639-024-12596-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYan, L. (2023). AI-assisted language learning and EFL writing performance. \u003cem\u003eFrontiers in Psychology\u003c/em\u003e, \u003cem\u003e14\u003c/em\u003e., Article 1261955. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fpsyg.2023.1261955\u003c/span\u003e\u003cspan address=\"10.3389/fpsyg.2023.1261955\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYuan, L., \u0026amp; Liu, X. (2025). The effect of artificial intelligence tools on EFL learners' engagement, enjoyment, and motivation. \u003cem\u003eComputers in Human Behavior\u003c/em\u003e, \u003cem\u003e162\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.chb.2024.108474\u003c/span\u003e\u003cspan address=\"10.1016/j.chb.2024.108474\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Article 108474.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYuksel, D., Soru\u0026ccedil;, A., Horzum, B., \u0026amp; McKinley, J. (2023). Examining the role of English language proficiency, language learning anxiety, and self-regulation skills in EMI students' academic success. \u003cem\u003eStudies in Second Language Learning and Teaching\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e(2), 399\u0026ndash;426. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.14746/ssllt.38280\u003c/span\u003e\u003cspan address=\"10.14746/ssllt.38280\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZagoto, I., Laia, R. D., Ndraha, L. D. M., \u0026amp; Wau, S. (2025). From policy to practice: a systematic review of English-medium instruction policy implementation across educational contexts. \u003cem\u003eCurrent Issues in Language Planning\u003c/em\u003e, 1\u0026ndash;17. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/14664208.2025.2598485\u003c/span\u003e\u003cspan address=\"10.1080/14664208.2025.2598485\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhai, Y., \u0026amp; Nezakatgoo, B. (2025). Evaluating AI-Powered Applications for Enhancing Undergraduate Students\u0026rsquo; Metacognitive Strategies, Self-Determined Motivation, and Social Learning in English Language Education. \u003cem\u003eScientific Reports\u003c/em\u003e, \u003cem\u003e15\u003c/em\u003e, 35199. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41598-025-19118-z\u003c/span\u003e\u003cspan address=\"10.1038/s41598-025-19118-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang, H., Liu, Y., Jiang, M., Chen, J., Wang, M., \u0026amp; Paas, F. (2025a). Emotional Artificial Intelligence in Education: A Systematic Review and Meta-Analysis. \u003cem\u003eEducational Psychology Review\u003c/em\u003e, \u003cem\u003e37\u003c/em\u003e(4), 106. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10648-025-10086-4\u003c/span\u003e\u003cspan address=\"10.1007/s10648-025-10086-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang, C., Schie\u0026szlig;l, J., Pl\u0026ouml;\u0026szlig;l, L., Hofmann, F., \u0026amp; Gl\u0026auml;ser-Zikuda, M. (2023). Acceptance of artificial intelligence among pre-service teachers: A multigroup analysis. \u003cem\u003eInternational Journal of Educational Technology in Higher Education\u003c/em\u003e, \u003cem\u003e20\u003c/em\u003e, 49. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s41239-023-00420-7\u003c/span\u003e\u003cspan address=\"10.1186/s41239-023-00420-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang, Q., Siraj, S. B., \u0026amp; Abdul Razak, R. B. (2025b). Effects of AI chatbots on EFL students' critical thinking skills and intrinsic motivation in argumentative writing. \u003cem\u003eInnovation in Language Learning and Teaching\u003c/em\u003e. Advance online publication. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/17501229.2025.2515111\u003c/span\u003e\u003cspan address=\"10.1080/17501229.2025.2515111\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"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":"asian-pacific-journal-of-second-and-foreign-language-education","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"jsfl","sideBox":"Learn more about [Asian-Pacific Journal of Second and Foreign Language Education](http://sfleducation.springeropen.com)","snPcode":"40862","submissionUrl":"https://submission.nature.com/new-submission/40862/3","title":"Asian-Pacific Journal of Second and Foreign Language Education","twitterHandle":"@SpringerOpen","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"network meta-analysis, artificial intelligence, English-Medium Instruction, learning motivation, computer-assisted language learning","lastPublishedDoi":"10.21203/rs.3.rs-8857786/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8857786/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eEnglish-Medium Instruction (EMI) programs have proliferated across global higher education, which create motivational obstacles for English as a Foreign Language (EFL) learners. These students frequently struggle with language barriers that hinder their engagement and self-confidence. Meanwhile, Artificial Intelligence (AI) technologies present innovative solutions, delivering customized and dynamic assistance to overcome such problems. To evaluate their impact, this research executed a network meta-analysis (NMA) on 15 empirical investigations, involving 1,847 university students in EMI settings. The analysis pitted three AI tools (i.e., Generative AI Chatbots, AI Writing Assistants, and AI Language Learning Applications) against conventional teaching methods to measure gains in learning motivation. Researchers employed standardized mean differences via Hedges' g for effect quantification, incorporating checks for heterogeneity and publication bias to ensure reliability. Findings demonstrated that every AI intervention markedly surpassed traditional approaches. Among them, AI Language Learning Applications achieved the strongest outcome at \u003cem\u003eg\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.907. Generative AI Chatbots trailed slightly behind with \u003cem\u003eg\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.892, while AI Writing Assistants registered a solid \u003cem\u003eg\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.692. Innovatively, the study uncovered how the responsive and flexible elements of chatbots and applications adeptly meet learners' needs for autonomy and competence, drawing from self-determination theory and promoting enduring motivation. Furthermore, subgroup evaluations showed no influence from variables like program length or student skill levels and implied widespread utility. Consequently, these insights furnish practical advice for embedding adaptive AI in EMI syllabi to heighten involvement and academic performance They also advocate for subsequent studies examining longitudinal effects and adaptations across cultures.\u003c/p\u003e","manuscriptTitle":"Impact of Artificial Intelligence Tools on Learning Motivation in University EMI Courses: A Network Meta-Analysis","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-23 18:55:36","doi":"10.21203/rs.3.rs-8857786/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-03-03T07:18:34+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-03-02T03:16:25+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"232999540237441421389236884737200057454","date":"2026-03-02T03:13:50+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-26T07:53:52+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-19T11:54:03+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"315200091076217269318900602359638119371","date":"2026-02-19T11:15:32+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"129633488504595500747697947788454777053","date":"2026-02-17T03:26:14+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-02-17T03:10:50+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-02-17T03:01:40+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-02-16T08:55:06+00:00","index":"","fulltext":""},{"type":"submitted","content":"Asian-Pacific Journal of Second and Foreign Language Education","date":"2026-02-12T05:23:05+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"asian-pacific-journal-of-second-and-foreign-language-education","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"jsfl","sideBox":"Learn more about [Asian-Pacific Journal of Second and Foreign Language Education](http://sfleducation.springeropen.com)","snPcode":"40862","submissionUrl":"https://submission.nature.com/new-submission/40862/3","title":"Asian-Pacific Journal of Second and Foreign Language Education","twitterHandle":"@SpringerOpen","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"f245621b-cf92-4971-adbe-fb4d9f38a829","owner":[],"postedDate":"February 23rd, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-04-14T01:38:47+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-23 18:55:36","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8857786","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8857786","identity":"rs-8857786","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}