Assessing the Quality of a Personalized Prompt Generator and AI-Chatbot (ChatGPT) for Dietary and Exercise Planning in Obese Adults Using the Fuzzy Delphi Method

Assessing the Quality of a Personalized Prompt Generator and AI-Chatbot (ChatGPT) for Dietary and Exercise Planning in Obese Adults Using the Fuzzy Delphi Method | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (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];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search Assessing the Quality of a Personalized Prompt Generator and AI-Chatbot (ChatGPT) for Dietary and Exercise Planning in Obese Adults Using the Fuzzy Delphi Method Azwa Suraya Mohd Dan , Adam Linoby , Sazzli Shahlan Kasim , Sufyan Zaki , Razif Sazali , Yusandra Yusoff , Zulqarnain Nasir , Amrun Haziq Abidin doi: https://doi.org/10.1101/2025.08.27.25334528 Azwa Suraya Mohd Dan 1 Faculty of Sports Science and Recreation, Universiti Teknologi MARA, Seremban Campus, Negeri Sembilan Branch , Negeri Sembilan, MALAYSIA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Adam Linoby 1 Faculty of Sports Science and Recreation, Universiti Teknologi MARA, Seremban Campus, Negeri Sembilan Branch , Negeri Sembilan, MALAYSIA Find this author on Google Scholar Find this author on PubMed Search for this author on this site For correspondence: linoby{at}uitm.edu.my Sazzli Shahlan Kasim 2 Faculty of Medicine, Universiti Teknologi MARA, Sungai Buloh Campus, Selangor Branch , Selangor, MALAYSIA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Sufyan Zaki 3 Faculty of Sports Science and Recreation, Universiti Teknologi MARA , Shah Alam Campus, Selangor Branch , Selangor, MALAYSIA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Razif Sazali 1 Faculty of Sports Science and Recreation, Universiti Teknologi MARA, Seremban Campus, Negeri Sembilan Branch , Negeri Sembilan, MALAYSIA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Yusandra Yusoff 1 Faculty of Sports Science and Recreation, Universiti Teknologi MARA, Seremban Campus, Negeri Sembilan Branch , Negeri Sembilan, MALAYSIA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Zulqarnain Nasir 1 Faculty of Sports Science and Recreation, Universiti Teknologi MARA, Seremban Campus, Negeri Sembilan Branch , Negeri Sembilan, MALAYSIA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Amrun Haziq Abidin 1 Faculty of Sports Science and Recreation, Universiti Teknologi MARA, Seremban Campus, Negeri Sembilan Branch , Negeri Sembilan, MALAYSIA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Abstract Full Text Info/History Metrics Preview PDF Abstract Background The potential of artificial intelligence (AI) to personalize dietary and exercise advice for obesity management is increasingly evident. However, the effectiveness and appropriateness of AI-generated recommendations hinge significantly on input quality and structured guidance. Despite growing interest, there remains a notable gap regarding a robust and validated prompt-generation mechanism designed explicitly for obesity-related lifestyle planning. Objectives This study aimed to evaluate and refine the quality of a personalized AI-driven framework (NExGEN-ChatGPT) for dietary and exercise prescriptions in obese adults, employing the Fuzzy Delphi Method (FDM) to capture and integrate expert consensus. Methods A multidisciplinary expert panel, comprising 21 professionals from nutrition, medicine, psychology, fitness, and AI domains, was engaged in this study. Using structured questionnaires, the experts systematically assessed and refined six primary constructs, further detailed into several evaluative elements, resulting in the consensus validation of 111 specific criteria. Results Findings identified critical consensus-driven standards essential for personalized, safe, and feasible obesity management through AI. Moreover, the study revealed prioritized criteria pivotal for maintaining practical relevance, safety, and high-quality personalized recommendations. Conclusions Consequently, this validated framework provides a substantial foundation for subsequent real-world application and further research, thereby enhancing the effectiveness, scalability, and individualization of obesity interventions leveraging AI. 1. Introduction Globally, obesity has emerged as a profound public health crisis, with its prevalence steadily escalating over the last several decades. Recent estimates indicate that more than 650 million adults are obese worldwide, marking a near-tripling since 1975. 1 Notably, adult obesity prevalence has exceeded 40% in countries like the United States, substantially increasing healthcare expenditures and economic burdens due to obesity-related conditions. It is projected that obesity-driven healthcare costs and productivity losses could surpass 3% of global GDP by 2030. 2 Traditional one-to-one dietary and exercise counseling, although clinically effective (resulted in 5–10% body weight reduction over six to twelve months) remains resource-intensive and economically burdensome. 3 Consequently, there is a compelling need for scalable alternatives capable of delivering personalized guidance to vast populations. Artificial Intelligence (AI)-driven conversational agents, such as ChatGPT, have recently attracted attention as viable solutions, offering personalized nutrition and physical activity coaching accessible 24/7, transcending human counselor availability constraints. 4 Behavior-change theories, including the Capability, Opportunity, Motivation-Behavior (COM-B) model and Social Cognitive Theory, underpin the development of these chatbots, enhancing their capacity to facilitate effective, scalable lifestyle interventions. 5 Empirical investigations assessing AI chatbot-generated dietary and exercise recommendations demonstrate promising alignment with established clinical guidelines. Recent comparative analyses suggest chatbot-generated weekly exercise, and caloric targets align closely with recommendations from authoritative bodies such as the World Health Organization and the American College of Sports Medicine. 6 These AI-generated dietary plans typically adhere to recommended macronutrient balances, indicating basic safety and appropriateness. User engagement studies further underscore the potential of chatbots in weight management, highlighting sustained interactions driven by their accessibility and personalized feedback capabilities, resulting in weight-loss outcomes comparable to traditional counseling over short durations. 7 , 8 Nevertheless, significant concerns persist regarding the accuracy and safety of chatbot-generated advice. Documented incidents reveal cases where health chatbots inadvertently provided unsafe dietary recommendations or misinformation, necessitating rigorous validation, clear safety protocols, and oversight mechanisms prior to extensive deployment. 9 , 10 Systematic reviews and meta-analyses examining conversational AI agents for lifestyle modification reveal modest yet positive impacts on diet, physical activity, and body weight among obese populations. Recent meta-analyses have shown users of health chatbots typically achieve approximately 2–3 kg greater weight loss over three months compared to minimal interventions. 11 Despite encouraging findings, considerable variability in study methodologies and limited long-term evaluations (typically under six months) present substantial knowledge gaps concerning the durability of behavior changes facilitated by AI interventions. 12 , 13 Furthermore, the refinement of chatbot responses through advanced prompt-engineering strategies has shown potential to enhance personalization and clinical alignment of AI-generated advice, yet structured methods for optimizing these interactions remain nascent and insufficiently validated within obesity management. 14 , 15 To systematically address these critical quality concerns in AI-generated dietary and exercise advice, expert consensus methods such as the Fuzzy Delphi Method (FDM) are increasingly employed. This method incorporates fuzzy logic to quantify expert agreement on essential criteria for health technology evaluations, effectively managing uncertainty in subjective judgments. 16 Prior applications of FDM in telehealth and mobile health app evaluations have successfully established robust, expert-derived quality frameworks. 17 , 18 Recently developed instruments, including the Quality Assessment of Medical AI (QAMAI) scale, highlight efforts toward standardized evaluations of medical AI; however, domain-specific criteria for nutrition and fitness recommendations are still lacking, particularly regarding personalization, nutritional adequacy, and safety tailored explicitly for obesity management. 19 The Nutrition and Exercise Planning Artificial Intelligence Prompt Generator (NExGEN) is an online platform specifically developed to generate personalized dietary and exercise planning prompts for integration with AI-driven chatbots. NExGEN is designed to support individualized health and weight management goals by systematically incorporates diverse user-specific factors such as nutritional needs, fitness levels, anthropometric data, medical history, socioeconomic status, geographic location, stress levels, sleep patterns, and motivational aspects into its framework. Utilizing both offline (Microsoft Excel) and online spreadsheet applications (Microsoft Excel Online and Google Sheets), NExGEN produces highly tailored prompts designed explicitly for large language models such as ChatGPT, thereby facilitating precise AI-generated recommendations. The integration between NExGEN and ChatGPT capitalizes on AI’s decision-making capabilities to deliver personalized nutrition and exercise plans, promoting user autonomy and sustained adherence to lifestyle changes. Figure 1 illustrates the architecture framework of the NExGEN–ChatGPT system integration. Figure 1: Architecture framework of NExGEN–ChatGPT system integration. This research aims to rigorously assess and refine a novel integrated framework combining the NExGEN with ChatGPT, employing the FDM to derive expert consensus on the quality of generated lifestyle plans. Through iterative refinement, this approach seeks to validate the practicality, personalization, and accuracy of AI-driven dietary and exercise recommendations, providing a scalable, expert-validated solution to enhance obesity management outcomes among adults. 2. Methods This study adopted a structured quantitative design, employing the FDM to systematically refine and validate the NExGEN-ChatGPT prompt generation framework for delivering personalized fitness and nutritional recommendations. The FDM was specifically selected due to its unique combination of iterative consensus-building typical of traditional Delphi, 20 enhanced by fuzzy set theory, which accommodates subjective expert opinions and uncertainties inherent in interdisciplinary fields such as nutrition, fitness, psychology, medicine, and AI. In this approach, experts’ qualitative judgments, initially captured using Likert scales, were translated into quantitative Triangular Fuzzy Numbers represented by minimum (m1), moderate (m2), and maximum (m3) values. This allowed for capturing the nuanced and varying levels of agreement among experts effectively, thereby ensuring the accuracy and robustness of consensus outcomes in refining the framework. 2.1 NExGEN-ChatGPT Design and Process Flow The initial phase of the NExGEN-ChatGPT prompt generation process begins with the participant completing a comprehensive initial survey. This survey collects extensive and specific personal data across multiple relevant areas, including nutritional habits, fitness status, anthropometric measurements, demographics, socioeconomic factors, medical and physiological health history, geographical location, sleep patterns, stress levels, and motivational influences. Comprising 111 targeted questions structured across 12 distinct elements, this survey aims to establish an individualized baseline that informs the personalization of fitness and nutritional plans through ChatGPT. Crucially, participants first define specific personal health objectives, including desired outcomes such as targeted weight loss, timelines, and the starting dates of their weight management programs ( Figure 2 ). Detailed demographic and physiological measurements (e.g., body fat percentage, lean body mass, and heart rate) are collected alongside tailored questions concerning medical history, sleep, and stress management practices, further ensuring that the resulting personalized plans comprehensively reflect the individual’s specific needs. Figure 2: Starting interface of NExGEN initial survey. Following survey completion, participants encounter the first critical step in the NExGEN system, designated as the ‘1st Prompt’. This phase integrates the personalized data captured from the initial survey with a pre-defined narrative context designed explicitly to optimize interactions with ChatGPT. Participants copy this combined information which consist of detailed personal attributes alongside structured instructions into ChatGPT (using o3-mini model; the highest reasoning model at the time of data collection), thus enabling the chatbot to function effectively as a personal trainer, nutritionist, and health advisor ( Figure 3 ). This carefully structured process ensures the chatbot’s responses remain consistent and informed throughout subsequent interactions, establishing a foundation for continuous, data-driven engagement tailored specifically to the individual’s evolving health journey. Figure 3: Example of the 1st prompt page in NExGEN (left) and transposed data in ChatGPT (right). Subsequent prompts further build upon this initial setup to address various aspects of personalized health management. The 2nd Prompt specifically targets lifestyle and behavioral modifications, offering detailed recommendations on diet, physical activities, and psychological well-being. The 3rd Prompt generates a customized weekly exercise plan tailored to the individual’s profile and goals, complete with locations, timings, and specific exercise instructions. The 4th Prompt develops a detailed weekly dietary plan, structured around personalized nutrition needs, accessible food choices, and clearly defined caloric intakes. Subsequently, the 5th Prompt provides a categorized grocery shopping list designed to support adherence to the dietary plan, specifying necessary ingredients and feasible substitutions. Finally, the 6th Prompt delivers an easy-to-follow meal preparation guide, featuring user-friendly recipes, portion guidance, cooking tips, and food safety advice, thereby equipping users with practical tools to effectively manage their nutrition and maintain consistent progress toward their goals. 2.2 Research Instrument The primary instrument utilized in this research was a comprehensive self-administered questionnaire developed specifically for the FDM evaluation process. This bilingual questionnaire, available in Malay and English, was meticulously structured to facilitate clear understanding and accurate responses from the expert panel. It encompassed six distinct constructs, consisting of 111 specific items categorized into twelve critical evaluative elements relevant to fitness and nutritional planning. These constructs included detailed questions addressing demographic characteristics, physiological measurements, dietary habits, physical activity levels, psychological well-being, and behavioral motivations. Reliability of the instrument was established through cognitive pilot, 21 achieving Cronbach’s alpha coefficients ranging from 0.714 to 0.891, thereby confirming the questionnaire’s internal consistency and suitability for expert. 22 2.3 Data Collection and Preliminary Processing The process of data collection and analysis is based on the implementation steps of FDM, as shown in Figure 4 . Ethical clearance was secured from the Universiti Teknologi MARA Institutional Review Board (REC/09/2024 (PG/FB/30)), and the study was prospectively registered with the UMIN Trials Registry (UMIN000053570). Each prospective expert was approached individually, provided with a study briefing, and asked to sign an informed-consent form before receiving the questionnaire. To accommodate varying schedules and continuing public-health restrictions, questionnaires were distributed either face-to-face in hard copy or electronically as password-protected PDFs sent via e-mail or WhatsApp. All respondents completed the survey independently without researcher guidance, ensuring that responses reflected authentic, uninfluenced judgments. Figure 4: Schematic diagram of FDM threshold. 2.3.1 Step 1: Selection of Experts A purposive sampling strategy guided the selection of 21 multidisciplinary experts, each holding a minimum of ten years’ professional. 23 The panel included seasoned nutritionists, exercise specialists, medical practitioners, clinical psychologists, and AI researchers affiliated with academia, public healthcare institutions, and private sectors across Malaysia. This selection approach ensured comprehensive expert coverage across relevant disciplines, providing diverse and highly specialized insights crucial for refining the NExGEN-ChatGPT framework. Table 1 shows the demographic information and experts involved in this study. View this table: View inline View popup Table 1: Experts’ demography. The expert panel comprised of accredited nutritionists (n = 5), exercise specialists (n = 5), medical practitioners (n = 5), accredited psychologists (n = 4), and AI-researchers (n = 2). The panel was diverse in terms of gender, with 11 males and 10 females participating. The ages of the experts ranged from 36 to 58 years, with a mean age of 47.3 ± 7.6 years. This demographic data was systematically collected at the beginning of the study to ensure a broad range of insights and experiences, essential for the FDM applied in refining the AI-driven health management tools. 2.3.2 Step 2: Questionnaire Development The development of the questionnaire followed a rigorous, iterative, and evidence-informed approach. Initially, extensive literature reviews were conducted to map relevant theoretical constructs in personalized nutrition and exercise planning. Semi-structured expert interviews were subsequently performed to identify emerging factors not fully covered in existing literature. Potential constructs encompass user interaction with fitness and nutrition programs, tailored behavioral modification recommendations, and engagement’s psychological effects. In developing the questionnaire, these constructs are dissected into quantifiable variables for examination via targeted questions. This step considers the operationalization of each variable in a manner that aligns with the study’s goals and is clear to the participating experts. The synthesized findings were then meticulously operationalized into precise, measurable questionnaire items. The final instrument balanced quantitative Likert-scale assessments with qualitative open-ended questions to obtain both measurable ratings and in-depth insights necessary for robust Delphi analysis. 2.3.3 Step 3: Data Collection Experts completed the finalized questionnaire within a specified two-week timeframe, facilitated by scheduled follow-ups to ensure timely submissions. This structured approach allowed for comprehensive yet independent expert evaluations, minimizing biases from researcher interventions. Following the acquisition of consent, the questionnaire was provided in both hardcopy format for face-to-face meetings and in a softcopy PDF format for those unable to meet in person. Responses were systematically collected and securely documented, ensuring consistency and integrity throughout the data collection phase. 2.3.4 Step 4: Data Analysis a) Fuzzy Delphi Scale Expert evaluations collected via the Likert scales were converted to Triangular Fuzzy Numbers using a seven-point fuzzy scale to enhance analytical granularity. This conversion process effectively quantified the subjective expert judgments, facilitating precise aggregation and subsequent. 24 , 25 The number of levels on the Fuzzy scale is represented by odd numbers. A higher Fuzzy ranking indicates greater accuracy of the obtained data. Figure 5 illustrates the mean triangular graph plotted against the Triangular values, specifically the three components of the Triangular Fuzzy Number. Figure 5: Flowchart of FDM implementation for NExGEN-ChatGPT framework criteria evaluation. b) Consensus Evaluation The data analysis focuses on the use of Triangular Fuzzy Numbers to determine the value of the Threshold (d), which is important for achieving a level of expert agreement, as emphasized by. 26 In line with, 27 The degree of expert agreement was rigorously assessed using the vertex method to calculate inter-expert distances, setting a stringent d of ≤ 0.20 for acceptance. The vertex method was utilized to calculate the distance between the statistic for reporting ranked responses averages, enabling precise measurement of consensus among the experts. The distances for each Fuzzy number m = (m1, m2, m3) and n = (n1, n2, n3) calculated using the formula: Criteria were further validated by requiring at least a 75% consensus rate among experts. This dual-threshold system ensured robust methodological rigor, significantly enhancing the validity of the consensus-driven outcomes. 27 , 28 c) Defuzzification The final analytical step involved defuzzification, employing the weighted average formula to convert aggregated fuzzy values into clear and actionable scores. Criteria achieving a defuzzified score (Amax; A) of 0.50 or higher were retained, reflecting adequate expert consensus and while those below the threshold were systematically excluded. 29 The significance of the Fuzzy score (A) is important in establishing the ranking and priority of the elements under scrutiny. To calculate the Fuzzy score value (A), the formula Amax = 1/4 x (a1 + 2am + a2) was applied. The α-cut value, serving as a critical benchmark, is computed as the median of ‘0’ and ‘1’, resulting in α-cut = (0 + 1)/2 = 0.5. If the calculated Fuzzy score is found to be below this α-cut threshold, the criteria was deemed unacceptable, as this indicates a lack of expert agreement. This comprehensive defuzzification ensured the framework’s practical applicability and precision, readying the refined NExGEN-ChatGPT prompt generation system for subsequent validation studies and practical implementation. 3. Results The findings are explained according to constructs, elements, and criteria compiled based on discussions with experts. Overall, all the criteria accepted for each construct and element met the requirements and conditions to pass the expert consensus of >75%, the value of Threshold (d) < 0.2 and α -cut is more than 0.5. 3.1 Construct 1: Goal Settings Construct 1 focuses on goal setting and consists of three elements and six criteria. The elements specify the desired outcome for weight management, the targeted weight change, and the timeline for achieving these goals. Findings from the FDM analysis are reported in a table that lists only the criteria that met the acceptance threshold. 3.1.1 Element: Desired Weight Management Outcomes Table 2 highlights collective judgment on each criterion’s importance and applicability for ‘Desired Weight Management Outcomes’. The most significant criterion identified is the thorough initial and ongoing assessment of physical activity, which achieved unanimous expert agreement (100%), a threshold value of 0.127, and a high fuzzy score of 0.859. This underscores consistent assessment as essential for program effectiveness. Ranking second, utilizing specific metrics to measure improvements received a 90.48% consensus and a fuzzy score of 0.821, reinforcing the necessity for measurable outcomes. Aligning physical activity habits with overall health and weight management goals ranked third (90.5%, fuzzy score 0.765), emphasizing alignment with broader health objectives. Effective tracking systems and participant feedback ranked fourth (95.24%, 0.762), highlighting supportive infrastructure’s role. Technology tools such as apps and wearables received lower but notable acceptance (76.19%, 0.665). Lastly, diverse physical activities catering to individual interests showed moderate consensus (81%, 0.667), emphasizing personalized approaches. View this table: View inline View popup Table 2. Accepted Criteria Position by Ranking for the Element of Desired Weight Management Outcomes. 3.1.2 Element: Target Weight Management Outcomes Table 3 presents the prioritized criteria for ‘Target Weight Management Outcomes’ derived through the FDM, emphasizing their importance and feasibility based on expert consensus percentages and fuzzy scores. The highest-ranked criterion, clearly and precisely defining target weight change goals, achieved unanimous expert consensus (100%) and a fuzzy score of 0.806, highlighting clarity and specificity as essential for effective weight management. The alignment of weight change goals with established health standards, such as BMI and body fat percentage, ranked second, receiving 90.48% consensus and a fuzzy score of 0.786, underscoring the need for evidence-based objectives. Goals tailored realistically to individual health and fitness status ranked third (76.2%, 0.724), reinforcing the value of practicality. Personalization based on age, gender, and lifestyle ranked fourth (76.19%, 0.656), further supporting individualized goal-setting approaches. The safety of weight change goals ranked fifth (80.95%, 0.640), while realistic timelines, despite a high consensus (90.5%), had a lower fuzzy score (0.576), suggesting timeline practicality is less foundational compared to other criteria. View this table: View inline View popup Table 3: Accepted Criteria Position by Ranking for the Element of Target Weight Management Outcomes. 3.1.3 Element: Timeline for Achieving Goals Table 4 presents criteria ranked through the FDM concerning timelines for ‘Timeline for Achieving Goals’, based on expert consensus and fuzzy scores, emphasizing their relative importance and practical relevance. The top-ranked criterion, assessing if timelines are realistic and achievable considering individual fitness levels and lifestyles, attained an 85.7% consensus and the highest fuzzy score of 0.725, indicating the importance of aligning timelines with personal capabilities. The second-ranked criterion, structuring timelines to sustain motivation and engagement, received identical consensus (85.71%) but a lower fuzzy score (0.595), underscoring motivation’s role throughout the goal process. Flexibility in adjusting goals and timeframes ranked third (80.95%, 0.562), reinforcing the need for adaptable planning. Fourth, incorporating incremental milestones for monitoring progress and adjustments scored 85.7% consensus and a fuzzy score of 0.557, highlighting structured checkpoints. Personalization of timelines based on age, gender, health, and lifestyle was fifth (85.71%, 0.530). Lastly, prioritizing health and safety by avoiding rapid changes ranked sixth (81%, 0.529), stressing cautious timeline planning. View this table: View inline View popup Table 4: Accepted Criteria Position by Ranking for the Element of Timeline for Achieving Goals. 3.2 Construct 2: Demographics and Physical Characteristics Construct 2 addresses demographic and physical characteristics. It comprises three elements: age; height and weight; and physiological indicators, specifically body-fat percentage and lean body mass. 3.2.1 Element: Age The results in Table 5 rank criteria concerning ‘Age’ in nutrition and exercise recommendations using the FDM. The top criterion, assessing whether strategies are safe across age groups given health risks, had the highest expert consensus at 95.24% and a fuzzy score of 0.681, emphasizing safety and risk mitigation. The second-ranked criterion addressed the clarity and appropriateness of recommendations for diverse age groups (95.2% consensus, 0.673 fuzzy score), highlighting the necessity of clear, adaptable guidelines. Adaptability to individual differences ranked third (80.95% consensus, 0.648 fuzzy score), underscoring personalized recommendations. Specific nutritional and exercise requirements for each age group ranked fourth (90.5% consensus, 0.603 fuzzy score). Consideration of developmental and physiological variations was fifth (85.7%, 0.576 fuzzy score), while meeting specific nutritional needs was sixth (85.71%, 0.511 fuzzy score), establishing foundational importance. View this table: View inline View popup Table 5. Accepted Criteria Position by Ranking for the Element of Age. 3.2.2 Element: Body Composition The results presented in Table 6 prioritize criteria for nutrition and exercise recommendations based on ‘Body Composition’ using the FDM. The top-ranked criterion, addressing whether recommendations align with specific Body Mass Index (BMI) categories, achieved unanimous expert consensus (100%) and the highest fuzzy score (0.681), highlighting the importance of BMI-targeted strategies. The second criterion, assessing the safety of strategies across different height and weight profiles, also received full consensus (100%) and a fuzzy score of 0.605, emphasizing safety considerations. Personalization according to individual height and weight ranked third (95.2%, fuzzy score 0.576), underscoring tailored approaches. Realistic and achievable weight management goals placed fourth (85.71%, fuzzy score 0.557), ensuring practicality. Addressing nutritional needs across body sizes was fifth (90.48%, fuzzy score 0.517), emphasizing dietary inclusivity. Lastly, suitability of exercise intensity and type ranked sixth (76.19%, fuzzy score 0.511), indicating lower relative emphasis. View this table: View inline View popup Table 6. Accepted Criteria Position by Ranking for the Element of Body Composition. 3.2.3 Element: Physiological Measurements Table 7 presents prioritized element related to ‘Physiological Measurements’. The highest-ranked criterion is evaluating whether physiological measurements ensure participant safety and health monitoring during nutrition and exercise programs. This criterion achieved an expert consensus of 85.7% and the top fuzzy score (0.659), highlighting participant safety. Ranking second, the criterion on personalized strategies based on individual physiological data also received 85.7% consensus (0.630 fuzzy score), emphasizing tailored interventions. Employing modern technologies to enhance physiological data collection, analysis, and feedback ranked third (90.48% consensus, 0.613 fuzzy score), underscoring technology’s role in data accuracy. Ensuring reliable and accurate measurement tools ranked fourth (95.2% consensus, fuzzy score 0.576). Adjusting fitness and nutrition plans according to changing physiological data was fifth (80.95%, 0.573), with tracking and reporting systems ranking sixth (80.95%, 0.511), reflecting their supporting roles. View this table: View inline View popup Table 7. Accepted Criteria Position by Ranking for the Element of Physiological Measurements. 3.3 Construct 3: Health and Medical History The ‘Health and Medical History’ is the third construct with 2 element and 6 criteria for each element. 3.3.1 Element: Medical Conditions Table 8 presents analysis results related to ‘Medical Conditions,’ emphasizing tailored and inclusive strategies for fitness and nutrition management. The criterion advocating comprehensive inclusion of diverse medical conditions received 81% expert consensus and the highest fuzzy score (0.741), emphasizing broad applicability. Ranked second, customizing exercise and diet plans for specific medical conditions obtained unanimous consensus (100%, fuzzy score 0.683), highlighting personalized adaptation. Accessibility and feasibility for individuals with medical limitations ranked third (85.71%, fuzzy score 0.630), reinforcing inclusivity. Fourth-ranked was the presence of feedback mechanisms to assess program effectiveness in managing medical conditions (90.48%, fuzzy score 0.613), emphasizing continuous improvement. Safety considerations for participants with varying medical conditions ranked fifth (85.7%, fuzzy score 0.594), underlining risk reduction. Lastly, monitoring the impact of fitness and nutrition programs on medical conditions ranked sixth (80.95%, fuzzy score 0.530), underscoring adaptability based on changing health conditions. View this table: View inline View popup Table 8. Accepted Criteria Position by Ranking for the Element of Medical Conditions. 3.2.3 Element: Medications Table 9 presents prioritized criteria related to the ‘Medications’ construct, emphasizing medication considerations in fitness and nutrition planning. The highest-ranked criterion which is ensuring exercise and nutrition recommendations avoid negative interactions with prescribed medications, achieved 95.2% consensus and the highest fuzzy score (0.698), underscoring the significance of safety. Second-ranked was effective customization of health plans based on medication effects (81% consensus, fuzzy score 0.667), highlighting personalized approaches. Adaptability of plans to accommodate medication regimen changes ranked third (76.2%, fuzzy score 0.663), emphasizing program flexibility. Fourth-ranked was responsiveness regarding necessary medication adjustments due to fitness and nutrition changes (95.24% consensus, fuzzy score 0.644). Ethical and legal considerations for medication-related privacy ranked fifth (85.71%, fuzzy score 0.622), reinforcing participant trust. Lastly, integrating medical records and medication histories into planning ranked sixth (80.95%, fuzzy score 0.529), indicating foundational yet secondary importance. Results emphasize safety, personalization, adaptability, and ethical compliance in health interventions. View this table: View inline View popup Table 9. Accepted Criteria Position by Ranking for the Element of Medications. 3.4 Construct 4: Lifestyle and Behaviour 3.4.1 Element: Sleep Pattern Table 10 highlights the element of ‘Sleep Pattern’. The highest-ranked criterion, adjusting programs for participants with diagnosed sleep disorders or irregular sleep patterns, achieved 90.48% consensus and a fuzzy score of 0.673, indicating strong agreement on addressing individual sleep-related challenges. Addressing the impact of sleep quality and duration on fitness and health outcomes ranked second (76.2% consensus, 0.665), reflecting sleep’s role in recovery, weight control, and energy. Educating participants on sleep’s connection to health ranked third (80.95%, 0.638). Integrating sleep patterns into personalized plans ranked fourth (81%, 0.637), emphasizing tailored approaches. Customized recommendations based on individual sleep quality ranked fifth (85.7%, 0.602), followed by monitoring systems for tracking sleep (85.71%, 0.583), highlighting data-driven approaches to improving health outcomes. View this table: View inline View popup Table 10. Accepted Criteria Position by Ranking for the Element of Sleep Pattern. 3.4.2 Element: Stress Level The findings from Table 11 shows the prioritized criteria for ‘Stress Level’. The top-ranked criterion, considering stress levels in relation to health outcomes, such as weight management, achieved the highest expert consensus (95.2%) and fuzzy score (0.784), highlighting stress as a significant factor affecting physical and mental health. Ranked second, conducting initial and ongoing stress assessments to tailor program recommendations obtained 81% consensus and a fuzzy score of 0.738, emphasizing personalized evaluation. Integrating stress management into fitness and nutrition programs ranked third (76.2%, 0.663). Ensuring the safety and effectiveness of stress management methods ranked fourth (80.95%, 0.621), followed by promoting relaxation techniques like meditation and yoga (80.95%, 0.538). Customizing stress management based on individual triggers ranked sixth (80.95%, 0.529), supporting a holistic, personalized approach to wellness. View this table: View inline View popup Table 11. Accepted Criteria Position by Ranking for the Element of Stress Level. 3.4.3 Element: Physical Activity Habits The results from Table 12 prioritize criteria for integrating ‘Physical Activity Habits’. The highest-ranked criterion, which is effectively aligning physical activity habits with overall health and weight management goals have achieved the top fuzzy score (0.781) and expert consensus (85.7%), emphasizing alignment with broader health objectives. The use of technology, such as apps and wearable devices, ranked second (85.71%, 0.760), highlighting its role in enhancing participant engagement. Consistent initial and ongoing assessments of physical activity levels ranked third (95.2%, 0.689), underscoring the importance of regular monitoring. Specific metrics to measure physical activity improvements ranked fourth (90.48%, 0.630), followed by tracking and feedback systems (80.95%, 0.519). Finally, incorporating diverse physical activities to accommodate varied interests ranked sixth (76.2%, 0.510), stressing inclusivity and sustained engagement. View this table: View inline View popup Table 12. Accepted Criteria Position by Ranking for the Element of Physical Activity Habits. 3.4.4 Element: Dietary Habits The results from Table 13 prioritize ‘Dietary Habits’ criteria. The highest-ranked criterion, consideration of environmental impacts and sustainability in dietary recommendations, achieved the top fuzzy score (0.768) and expert consensus (90.48%), highlighting alignment with ecological sustainability. Ranked second, aligning dietary habits with individual fitness and weight management goals scored 0.700 with a consensus of 76.2%, emphasizing personalized dietary integration. Thorough assessments of dietary habits during program initiation and progression ranked third (100%, 0.689), underscoring continuous monitoring. Fourth was the consideration of dietary habits’ psychological effects (95.24%, 0.684), followed by ensuring dietary recommendations’ safety based on medical history (81%, 0.648). Lastly, technology use, such as meal-tracking apps, ranked sixth (85.71%, 0.511), enhancing adherence and participant engagement. View this table: View inline View popup Table 13. Accepted Criteria Position by Ranking for the Element of Dietary Habits. 3.5 Construct 5: Motivation and Social Support 3.5.1 Element: Motivations for Weight Management Table 14 highlights the analysis of ‘Motivations for Weight Management’ highlights expert consensus on tailored strategies. Regular feedback and positive reinforcement were ranked highest (fuzzy score 0.700; 76.19% consensus), emphasizing continuous feedback for sustained motivation. Alignment of strategies with individual health goals ranked second (0.698; 100% consensus), reinforcing goal-oriented personalization. Third was tailoring strategies to personal backgrounds (0.665; 76.2%), followed by the clarity of motivational relevance to specific goals (0.640; 85.7%), integration into daily routines (0.592; 85.71%), and adaptability based on changing needs (0.529; 80.95%). These findings support structured, individualized, and adaptable motivational frameworks. View this table: View inline View popup Table 14. Accepted Criteria Position by Ranking for the Element of Motivations for Weight Managements. 3.6 Construct 6: Environmental Factors 3.6.1 Element: Environmental Factors The analysis of grocery shopping habits within “Environmental Factors” highlights their importance in supporting nutritional strategies ( Table 15 ). Personalized shopping strategies aligned with individual dietary preferences ranked highest (fuzzy score 0.768; 90.5% consensus). Feedback mechanisms evaluating shopping habit changes ranked second (0.668; 95.2%), emphasizing regular assessment. Alignment of grocery habits with program nutritional goals ranked third (0.640; 81%), followed by integration with meal planning (0.529; 81%), educational programs promoting healthy choices (0.517; 90.5%), and the influence of shopping behavior on dietary outcomes (0.510; 76.2%). Results indicate personalized, supported, education-oriented grocery strategies enhance dietary success. View this table: View inline View popup Table 15. Accepted Criteria Position by Ranking for the Element of Environmental Factors. 4. Discussion The findings present expert consensus outcomes derived through the FDM. All criteria related to constructs and elements were validated based on agreement among multidisciplinary experts in obesity management, AI, nutrition, psychology, and fitness. This expert consensus informed the development of an integrated framework using the NExGEN-ChatGPT prompt generator for personalized dietary and exercise planning for obese adults. The framework encompasses six primary constructs, subdivided into distinct evaluative elements, and a total of 111 specific criteria, each meeting stringent FDM acceptance standards (>75% expert consensus, threshold value (d) 0.5). These validated criteria collectively form a structured hierarchy model illustrating the integrated AI-based obesity management framework presented in subsequent sections. The findings from this study provide significant insights into the framework for personalized diet and exercise planning through the NExGEN-ChatGPT system, validated by expert consensus using the FDM. The expert panel identified and prioritized six key constructs, broken down into several elements and criteria, to evaluate the quality of AI-generated plans. These criteria, which met the thresholds of >75% consensus, d 0.5, offer a robust basis for developing and evaluating AI-driven weight management tools. Expert-derived hierarchical models, such as those used in digital-health interventions, 30 provide a structured approach that helps ensure that the most important quality dimensions (e.g., personalization, safety, and accuracy) are prioritized in system design. Notably, a comparison with real-world outputs reveals a gap between idealized expert models and the actual performance of ChatGPT when generating dietary and exercise plans. 31 This gap emphasizes the importance of continuous iteration and refinement of the prompt generator to better align AI outputs with expert expectations. Overall, the hierarchical model derived from expert consensus offers a meaningful step towards improving AI integration into personalized obesity care, although further development is needed to ensure consistency with expert recommendations. The study further highlights the importance of multi-construct frameworks in AI-driven obesity interventions. The top-10 prioritized criteria derived from expert consensus include critical factors such as personalization depth, clarity of content, goal alignment, and safety. These criteria directly correspond to established behavior change theories, such as Goal-Setting Theory and Self-Determination Theory, 32 which emphasize that goal specificity, user engagement, and empathy significantly influence weight loss success. Research has shown that personalized, clear, and evidence-based recommendations consistently improve the accuracy of chatbot responses, particularly in health domains like nutrition and. 33 Furthermore, by benchmarking AI-generated diet and exercise plans against clinical guidelines (i.e., from the American College of Sports Medicine and the American Diabetes Association) the study found that while ChatGPT’s advice aligns generally with these standards, certain element like adapting advice to individual medical conditions are occasionally overlooked. 34 This underscores the necessity of refining the prompt generator to accommodate more specific health profiles and conditions, ensuring that the AI can consistently deliver highly personalized, medically appropriate guidance. Goal-setting is a central component of successful obesity management, as emphasized by the results of this study, which align with findings from prior research. 35 Experts highlighted the importance of defining clear, specific, and achievable goals (i.e., based on the SMART criteria) within the NExGEN-ChatGPT framework. The use of specific and measurable targets has been shown to significantly improve engagement and weight-loss success in obese adults. 36 The study demonstrated that integrating goal-setting with continuous feedback, such as tracking steps via wearables or app-based systems, enhances adherence to diet and exercise plans. These real-time tracking mechanisms are essential in reinforcing goal attainment and adjusting goals dynamically based on user progress. However, as noted in previous studies, there are safety concerns when setting overly ambitious goals for users with specific medical conditions, such as those with joint problems or cardiovascular risks. 37 By integrating expert feedback into the goal-setting process, the study ensures that targets are both safe and achievable, fostering a sense of accomplishment that promotes long-term adherence. The findings also support the growing body of research demonstrating the effectiveness of wearables and mobile apps in promoting physical activity among obese adults. Consistent with studies, 38 the NExGEN-ChatGPT system, when paired with wearable technology, provided a comprehensive tool for tracking activity levels and adjusting exercise recommendations in real-time. The incorporation of technology such as fitness trackers (Fitbit, Garmin) has been linked to improved user engagement, as users are motivated by continuous feedback and real-time reminders to stay active. 39 Additionally, the study found that personalized, empathetic responses from the chatbot during moments of low motivation or setbacks helped users persist in their exercise routines. This approach aligns with research highlighting the critical role of empathetic, non-judgmental communication in maintaining long-term engagement with weight-loss programs. 33 In contrast, conflict-prone interactions, where users perceive the AI as overly strict or unempathetic, may lead to disengagement and a decline in physical activity. Therefore, fostering a positive, supportive chatbot-user relationship is essential to sustaining user engagement and ensuring effective behavior change. The importance of safety and medical sensitivity in AI-generated diet and exercise plans was underscored in the expert consensus. As with earlier research, 40 the study emphasizes that chatbots must be equipped with robust safeguards to avoid contraindicated recommendations, particularly for users with comorbidities such as diabetes, hypertension, or heart disease. Expert consensus around safety filters and medical history integration was critical in shaping the NExGEN-ChatGPT system, ensuring that recommendations were not only personalized but also medically appropriate. Previous studies have highlighted instances where AI systems, including ChatGPT, delivered potentially harmful advice, especially when users’ medical conditions were not fully considered. 34 To address this, the study incorporated expert-designed rules for medication-nutrition and medication-exercise interactions, reinforcing the importance of ensuring that AI outputs are aligned with clinical guidelines and best practices. Moreover, privacy and security were prioritized, as safeguarding user health data is essential for fostering trust in AI systems. 36 These safeguards ensure that the NExGEN-ChatGPT system delivers safe, reliable, and ethical health advice, adhering to both regulatory standards and expert recommendations. The study also highlights the significance of incorporating lifestyle factors such as sleep, stress, and behavioral habits into personalized obesity management. Research consistently shows that inadequate sleep and chronic stress can undermine weight-loss efforts, with insufficient sleep associated with higher weight regain and poorer adherence to exercise. 35 The NExGEN-ChatGPT system’s ability to assess and provide tailored advice on these factors is consistent with the holistic approach advocated in recent obesity interventions. By incorporating behavioral health strategies and offering personalized feedback, the chatbot not only addresses physical activity and dietary behaviors but also promotes mental well-being, reducing stress and enhancing motivation. Furthermore, the study aligns with the use of adaptive algorithms in AI systems, where real-time data on sleep, stress, and activity levels enable dynamic adjustments to the user’s plan. 37 This ability to continuously adapt and provide relevant, personalized support is a key feature of AI-driven obesity interventions and plays a pivotal role in ensuring long-term success. The findings of this study contribute to the growing body of research on AI-driven obesity management by presenting a robust, expert-validated framework for personalized dietary and exercise plans using the NExGEN-ChatGPT system. This system, grounded in expert consensus, offers a promising tool for enhancing weight management through tailored recommendations, yet its efficacy in real-world settings remains to be fully explored. Future studies should aim to test the NExGEN-ChatGPT system’s effectiveness in diverse populations, including those with varying comorbidities, across different cultural contexts, and with long-term follow-up to assess sustained behavior change and weight loss outcomes. While this study provides valuable insights, its limitations include the reliance on expert consensus without direct user trials, the potential for bias in self-reported data, and the absence of diverse socio-economic and ethnic representation in the expert panel. Further research is needed to validate these findings through randomized controlled trials and to refine the system’s adaptability for broader, more inclusive user groups. 5. Conclusion This study has successfully developed and validated a personalized prompt generation framework using the NExGEN-ChatGPT system, informed by expert consensus through the FDM analysis. The findings highlight the significance of integrating expert-derived quality criteria, such as personalization, safety, and clarity, in the design of AI-driven obesity management tools. The expert panel identified and prioritized critical constructs, which were used to tailor the system’s dietary and exercise recommendations to individual needs. These results demonstrate the potential for AI to support weight management interventions, offering a scalable and accessible alternative to traditional methods. However, while the system’s framework shows promise, the study’s limitations which may include the reliance on expert consensus and the lack of real-world testing obviously underscore the need for further validation. Future research should test the NExGEN-ChatGPT system in diverse populations and real-world settings to assess its long-term efficacy, particularly in terms of sustained weight loss and behavioral change. Additionally, refining the system’s adaptability to individual medical conditions and cultural contexts will be crucial to ensuring its widespread applicability. Overall, this study lays the groundwork for the development of AI-assisted, personalized obesity management tools that can enhance treatment accessibility and effectiveness, providing a valuable tool in the global fight against obesity. Declarations Author contribution statement Azwa Suraya Mohd Dan and Adam Linoby: Conceived and designed the experiments; Performed the experiments; Analyzed and interpreted the data; Contributed to data analysis and interpretation of the data; Wrote the paper. Sazzli Shahlan Kasim, Sufyan Zaki, and Razif Sazali: Conceived and designed the study; Interpreted the data. Yusandra Md Yusoff, Zulqarnain Nasir, and Amrun Haziq Abidin: Contributed to data analyses. Disclosure of funding No funding was received for this study. Data availability statement Data will be made available on request. Conflict of interests Azwa Suraya Mohd Dan, Adam Linoby, Sazzli Shahlan Kasim, Sufyan Zaki, Razif Sazali, Yusandra Md Yusoff, Zulqarnain Nasir, and Amrun Haziq Abidin declare that they have no conflict of interest. Acknowledgment The authors would like to express my heartfelt gratitude to all the experts who contributed their valuable insights to this study. Their expertise in nutrition, exercise, medicine, psychology, and AI has been integral to the development of the NExGEN-ChatGPT framework for personalized weight management. All authors read and approved the final manuscript. References 1. ↵ Lara FVO de , Fernández CT , Zurita-Campos J , Hernández JAM . Obesity and overweight . Medicine (Spain) . 2024 ; 14 ( 15 ). doi: 10.1016/j.med.2024.08.001 OpenUrl CrossRef 2. ↵ Lobstein T , Jackson-Leach R , Powis J , Brinsden H , Gray M. World Obesity Atlas 2023 . World Obesity Federation . 2023 ;(March). 3. ↵ Wharton S , Lau D , Vallis M , et al. Obesity in adults: A clinical practice guideline . CMAJ . 2020 ; 192 ( 31 ). doi: 10.1503/cmaj.191707 OpenUrl FREE Full Text 4. ↵ Milne-Ives M , de Cock C , Lim E , et al. The effectiveness of artificial intelligence conversational agents in health care: Systematic review . J Med Internet Res . 2020 ; 22 ( 10 ): e20346 . doi: 10.2196/20346 OpenUrl CrossRef PubMed 5. ↵ Hauser-Ulrich S , Künzli H , Meier-Peterhans D , Kowatsch T . A smartphone-based health care chatbot to promote self-management of chronic pain (SELMA): Pilot randomized controlled trial . JMIR Mhealth Uhealth . 2020 ; 8 ( 4 ). doi: 10.2196/15806 OpenUrl CrossRef PubMed 6. ↵ Ponzo V , Rosato R , Scigliano MC , et al. Comparison of the accuracy, completeness, reproducibility, and consistency of different AI chatbots in providing nutritional advice: An exploratory study . J Clin Med . 2024 ; 13 ( 24 ). doi: 10.3390/jcm13247810 OpenUrl CrossRef 7. ↵ Forman EM , Berry MP , Butryn ML , et al. Using artificial intelligence to optimize delivery of weight loss treatment: Protocol for an efficacy and cost-effectiveness trial . Contemp Clin Trials . 2023 ; 124 . doi: 10.1016/j.cct.2022.107029 OpenUrl CrossRef 8. ↵ Oh Y , Zhang J , Fang M , Fukuoka Y . A systematic review of artificial intelligence chatbots for promoting physical activity, healthy diet, and weight loss . International Journal of Behavioral Nutrition and Physical Activity . 2021 ; 18 ( 1 ). doi: 10.1186/s12966-021-01224-6 OpenUrl CrossRef 9. ↵ Sharp G , Torous J , West ML . Ethical challenges in AI approaches to eating disorders . J Med Internet Res . 2023 ; 25 . doi: 10.2196/50696 OpenUrl CrossRef 10. ↵ Meyrowitsch DW , Jensen AK , Sørensen JB , Varga T V . AI chatbots and (mis)information in public health: impact on vulnerable communities . Front Public Health . 2023 ; 11 . doi: 10.3389/fpubh.2023.1226776 OpenUrl CrossRef 11. ↵ Zhang J , Oh YJ , Lange P , Yu Z , Fukuoka Y . Artificial intelligence chatbot behavior change model for designing artificial intelligence chatbots to promote physical activity and a healthy diet: Viewpoint . J Med Internet Res . 2020 ; 22 ( 9 ). doi: 10.2196/22845 OpenUrl CrossRef 12. ↵ Ufholz K , Bhargava D . A review of telemedicine interventions for weight loss . Curr Cardiovasc Risk Rep . 2021 ; 15 ( 9 ). doi: 10.1007/s12170-021-00680-w OpenUrl CrossRef 13. ↵ Singh B , Olds T , Brinsley J , et al. Systematic review and meta-analysis of the effectiveness of chatbots on lifestyle behaviours . NPJ Digit Med . 2023 ; 6 ( 1 ). doi: 10.1038/s41746-023-00856-1 OpenUrl CrossRef 14. ↵ Yang Z , Khatibi E , Nagesh N , et al. ChatDiet: Empowering personalized nutrition-oriented food recommender chatbots through an LLM-augmented framework . Smart Health . 2024 ; 32 : 100465 . doi: 10.1016/J.SMHL.2024.100465 OpenUrl CrossRef 15. ↵ Wang J , Shi E , Yu S , et al. Prompt engineering for healthcare: methodologies and applications . Published online April 28 , 2023 . http://arxiv.org/abs/2304.14670 16. ↵ Jakubczyk M , Kamiński B . Fuzzy approach to decision analysis with multiple criteria and uncertainty in health technology assessment . Ann Oper Res . 2017 ; 251 ( 1-2 ). doi: 10.1007/s10479-015-1910-9 OpenUrl CrossRef 17. ↵ Chew HSJ , Chew NW , Loong SSE , et al. Effectiveness of a novel artificial intelligence-assisted weight loss app on improving eating behaviors: A mixed-method evaluation . Obes Res Clin Pract . 2023 ; 17 ( 5 ). doi: 10.1016/j.orcp.2023.08.015 OpenUrl CrossRef 18. ↵ Manoliadis OG , Pantouvakis JP , Christodoulou S . Improving qualifications-based selection by use of the fuzzy Delphi method . Construction Management and Economics . 2009 ; 27 ( 4 ). doi: 10.1080/01446190902758993 OpenUrl CrossRef 19. ↵ Vaira LA , Lechien JR , Abbate V , et al. Validation of the Quality Analysis of Medical Artificial Intelligence (QAMAI) tool: a new tool to assess the quality of health information provided by AI platforms . European Archives of Oto-Rhino-Laryngology . Published online November 1, 2024 . doi: 10.1007/s00405-024-08710-0 OpenUrl CrossRef 20. ↵ Yin TT , Hanif H . Fuzzy delphi method: A step-by-step guide to obtain expert consensus on MUETBot functionalities . International Journal of Academic Research in Business and Social Sciences . 2024 ; 14 ( 4 ). doi: 10.6007/ijarbss/v14-i4/21307 OpenUrl CrossRef 21. ↵ Schwarz N , Oyserman D . Asking questions about behavior: cognition, communication, and questionnaire construction . Am J Eval . 2001 ; 22 ( 2 ): 127 – 160 . doi: 10.1016/S1098-2140(01)00133-3 OpenUrl CrossRef 22. ↵ Tavakol M , Dennick R . Making sense of Cronbach’s alpha . Int J Med Educ . 2011 ; 2 . doi: 10.5116/ijme.4dfb.8dfd OpenUrl CrossRef PubMed 23. ↵ Keeney S , Hasson F , Mckenna H . The Delphi Technique in Nursing and Health Research .; 2010 . doi: 10.1002/9781444392029 OpenUrl CrossRef 24. ↵ Pervez H , Ali Y , Pamucar D , Garai-Fodor M , Csiszárik-Kocsir Á . Evaluation of critical risk factors in the implementation of modular construction . PLoS One . 2022 ; 17 (8 August). doi: 10.1371/journal.pone.0272448 OpenUrl CrossRef 25. ↵ Mohamed S , Ismail NK , Hamzah MI . The application of fuzzy delphi technique in the aspect of school connectedness as preventive measure of preschool children delinquency . International Journal of Academic Research in Progressive Education and Development . 2019 ; 8 ( 4 ). doi: 10.6007/ijarped/v8-i4/6783 OpenUrl CrossRef 26. ↵ Thomaidis , Nikitakos , Dounias. The evaluation of information technology projects: A fuzzy multicriteria decision-making approach . Int J Inf Technol Decis Mak . 2006 ; 5 ( 1 ). doi: 10.1142/S0219622006001897 OpenUrl CrossRef 27. ↵ Cheng CH , Lin Y . Evaluating the best main battle tank using fuzzy decision theory with linguistic criteria evaluation . Eur J Oper Res . 2002 ; 142 ( 1 ): 174 – 186 . doi: 10.1016/S0377-2217(01)00280-6 OpenUrl CrossRef Web of Science 28. ↵ Barrett D , Heale R . What are Delphi studies? Evid Based Nurs . 2020 ; 23 ( 3 ). doi: 10.1136/ebnurs-2020-103303 OpenUrl FREE Full Text 29. ↵ Morim A , Sa Fortes E , Reis P , Cosenza C , Doria F , Goncalves A . Think Fuzzy System : Developing New Pricing Strategy Methods for Consumer Goods Using Fuzzy Logic . International Journal of Fuzzy Logic Systems . 2017 ; 7 ( 1 ). doi: 10.5121/ijfls.2017.7101 OpenUrl CrossRef 30. ↵ Denecke K , Abd-Alrazaq A , Househ M , Warren J . Evaluation Metrics for Health Chatbots: A Delphi Study . Methods Inf Med . 2021 ; 60 ( 5-6 ). doi: 10.1055/s-0041-1736664 OpenUrl CrossRef 31. ↵ Koopman B , Zuccon G . Dr ChatGPT tell me what I want to hear: How different prompts impact health answer correctness . In: EMNLP 2023 - 2023 Conference on Empirical Methods in Natural Language Processing, Proceedings . ; 2023 . doi: 10.18653/v1/2023.emnlp-main.928 OpenUrl CrossRef 32. ↵ Krukowski RA , Tilford JM , Harvey-Berino J , West DS . Comparing behavioral weight loss modalities: Incremental cost-effectiveness of an internet-based versus an in-person condition . Obesity . 2011 ; 19 ( 8 ). doi: 10.1038/oby.2010.341 OpenUrl CrossRef PubMed 33. ↵ Rahmanti AR , Yang HC , Bintoro BS , et al. SlimMe, a Chatbot With Artificial Empathy for Personal Weight Management: System Design and Finding . Front Nutr . 2022 ; 9 . doi: 10.3389/fnut.2022.870775 OpenUrl CrossRef 34. ↵ Walker HL , Ghani S , Kuemmerli C , et al. Reliability of medical information provided by ChatGPT: Assessment against clinical guidelines and patient information quality instrument . J Med Internet Res . 2023 ; 25 . doi: 10.2196/47479 OpenUrl CrossRef PubMed 35. ↵ Wren GM , Koutoukidis DA , Scragg J , Whitman M , Jebb S . The association between goal setting and weight Loss: prospective analysis of a community weight loss program . J Med Internet Res . 2023 ; 25 : e43869 . doi: 10.2196/43869 OpenUrl CrossRef PubMed 36. ↵ Lehmann M , Jones L , Schirmann F . App Engagement as a Predictor of Weight Loss in Blended-Care Interventions: Retrospective Observational Study Using Large-Scale Real-World Data . J Med Internet Res . 2024 ; 26 . doi: 10.2196/45469 OpenUrl CrossRef 37. ↵ Patra E , Kokkinopoulou A , Wilson-Barnes S , et al. Personal goals, user engagement, and meal adherence within a personalised AI-Based mobile application for nutrition and physical activity . Life . 2024 ; 14 ( 10 ). doi: 10.3390/life14101238 OpenUrl CrossRef 38. ↵ Chew HSJ . The use of Artificial Intelligence-based conversational agents (Chatbots) for weight loss: Scoping review and practical recommendations . JMIR Med Inform . 2022 ; 10 ( 4 ). doi: 10.2196/32578 OpenUrl CrossRef PubMed 39. ↵ Robles N , Puig EP , Gómez-Calderón C , et al. Evaluation criteria for weight management apps: Validation using a modified delphi process . JMIR Mhealth Uhealth . 2020 ; 8 ( 7 ). doi: 10.2196/16899 OpenUrl CrossRef 40. ↵ Sharp G , Torous J , West ML . Ethical challenges in AI approaches to eating disorders . J Med Internet Res . 2023 ; 25 . doi: 10.2196/50696 OpenUrl CrossRef View the discussion thread. Back to top Previous Next Posted August 28, 2025. Download PDF Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Assessing the Quality of a Personalized Prompt Generator and AI-Chatbot (ChatGPT) for Dietary and Exercise Planning in Obese Adults Using the Fuzzy Delphi Method Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Assessing the Quality of a Personalized Prompt Generator and AI-Chatbot (ChatGPT) for Dietary and Exercise Planning in Obese Adults Using the Fuzzy Delphi Method Azwa Suraya Mohd Dan , Adam Linoby , Sazzli Shahlan Kasim , Sufyan Zaki , Razif Sazali , Yusandra Yusoff , Zulqarnain Nasir , Amrun Haziq Abidin medRxiv 2025.08.27.25334528; doi: https://doi.org/10.1101/2025.08.27.25334528 Share This Article: Copy Citation Tools Assessing the Quality of a Personalized Prompt Generator and AI-Chatbot (ChatGPT) for Dietary and Exercise Planning in Obese Adults Using the Fuzzy Delphi Method Azwa Suraya Mohd Dan , Adam Linoby , Sazzli Shahlan Kasim , Sufyan Zaki , Razif Sazali , Yusandra Yusoff , Zulqarnain Nasir , Amrun Haziq Abidin medRxiv 2025.08.27.25334528; doi: https://doi.org/10.1101/2025.08.27.25334528 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Health Informatics Subject Areas All Articles Addiction Medicine (568) Allergy and Immunology (863) Anesthesia (297) Cardiovascular Medicine (4421) Dentistry and Oral Medicine (443) Dermatology (382) Emergency Medicine (606) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1507) Epidemiology (15212) Forensic Medicine (30) Gastroenterology (1121) Genetic and Genomic Medicine (6581) Geriatric Medicine (667) Health Economics (996) Health Informatics (4520) Health Policy (1366) Health Systems and Quality Improvement (1611) Hematology (539) HIV/AIDS (1264) Infectious Diseases (except HIV/AIDS) (15906) Intensive Care and Critical Care Medicine (1103) Medical Education (620) Medical Ethics (144) Nephrology (667) Neurology (6580) Nursing (345) Nutrition (998) Obstetrics and Gynecology (1141) Occupational and Environmental Health (956) Oncology (3324) Ophthalmology (970) Orthopedics (369) Otolaryngology (420) Pain Medicine (435) Palliative Medicine (129) Pathology (663) Pediatrics (1689) Pharmacology and Therapeutics (691) Primary Care Research (710) Psychiatry and Clinical Psychology (5433) Public and Global Health (9212) Radiology and Imaging (2193) Rehabilitation Medicine and Physical Therapy (1368) Respiratory Medicine (1194) Rheumatology (593) Sexual and Reproductive Health (709) Sports Medicine (529) Surgery (709) Toxicology (99) Transplantation (288) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'9ff4e6980d6a0704',t:'MTc3OTM4MDM0NQ=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();