The application of large language models in bariatric surgery: A scoping review.

The application of large language models in bariatric surgery: A scoping review - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Search Log in Primary site navigation Search Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice PLoS One . 2026 Jun 5;21(6):e0350748. doi: 10.1371/journal.pone.0350748 The application of large language models in bariatric surgery: A scoping review Ningjing Guo 1, # , Xuyan Li 1, 2, # , Xiaoxue Li 1, ‡ , Congmin Kang 1, ‡ , Xiaoyan Gong 1, ‡ , Xinyu Ji 1, ‡ , Jie Zheng 1, * Editor: Hongyang Ma 3 Author information Article notes Copyright and License information 1 School of Nursing, Shanxi Medical University, Shanxi, China 2 Heping Hospital Affiliated to Changzhi Medical College, Shanxi, China 3 Peking University, CHINA ✉ * E-mail: zhengjie@sxmu.edu.cn Competing Interests: The authors have declared that no competing interests exist. # ☯ These authors contributed equally to this work and should be considered as co-first authors. ‡ These authors also contributed equally to this work. # Contributed equally. Roles Ningjing Guo : Data curation, Writing – original draft, Writing – review & editing Xuyan Li : Data curation, Writing – original draft, Writing – review & editing Xiaoxue Li : Data curation Congmin Kang : Data curation Xiaoyan Gong : Data curation Xinyu Ji : Data curation Jie Zheng : Data curation, Writing – review & editing Hongyang Ma : Editor Received 2026 Mar 9; Accepted 2026 May 18; Collection date 2026. © 2026 Guo et al This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. PMC Copyright notice PMCID: PMC13240874  PMID: 42247409 Abstract Background Exploratory applications of large language models within the specialized field of metabolic and bariatric surgery have begun to emerge. Nevertheless, existing research remains fragmented, lacking comprehensive integration. Objective To conduct a scoping review of studies on the application of large language models in the field of metabolic and bariatric surgery, aiming to provide a reference for clinical practice and future research. Methods This scoping review adhered to the Joanna Briggs Institute methodological framework and followed the preferred reporting items for systematic reviews and meta-Analyses extension for scoping reviews (PRISMA-ScR) guidelines.PubMed, Web of Science, The Cochrane Library, Embase, CINAHL, CNKI, Wanfang, and VIP databases were searched for relevant studies, with the search timeframe from database inception to November 2025. The included literature was summarized and analyzed. Results A total of 21 English-language studies were included. LLMs were primarily applied in scenarios such as patient education and information consultation, clinical decision support, and professional knowledge assessment. While LLMs performed well in information-provision tasks, they showed low consistency with expert opinions in complex clinical tasks such as individualized surgical recommendations. Performance varied across different models, with GPT-4 generally demonstrating superior performance, and domain-specific models showing professional potential. Current research still faces challenges regarding information accuracy, readability, and clinical applicability. Conclusion Large language models hold auxiliary potential in the field of metabolic and bariatric surgery, particularly for knowledge dissemination and patient education. However, their reliability in complex clinical decision-making remains limited. Future efforts should focus on conducting high-quality studies, advancing model specialization and standardized evaluation, and exploring safe and effective human-AI collaboration models. Introduction The global prevalence of obesity continues to rise [ 1 ]. Obesity and its associated medical problems, such as type 2 diabetes, hypertension, and metabolic dysfunction-associated steatohepatitis (MASH), have emerged as major public health challenges that impact population health and increase socioeconomic and healthcare burdens [ 2 , 3 ]. Metabolic and bariatric surgery (MBS) has become an effective intervention for severe obesity, demonstrating not only significant weight loss but also improvement in metabolic parameters and a reduction in the incidence of obesity-related diseases. It is currently one of the most effective approaches for treating severe obesity and related metabolic disorders [ 4 ]. In recent years, MBS has been widely adopted in China and entered a phase of rapid development. According to statistics from the Chinese Obesity and Metabolic Surgery Database, the total annual number of MBS procedures in China increased to approximately 37,249 in 2025 [ 5 ] reflecting the growing clinical demand and vitality of the specialty. However, alongside the sustained growth in surgical volume and the ongoing standardization of the specialty, metabolic and bariatric surgery continues to face a series of clinical challenges and practical pressures. For instance, in patient selection and surgical decision-making, choosing the most appropriate surgical technique remains a complex and contentious process [ 6 ]. Furthermore, due to the complexity of bariatric surgery and its long-term outcomes, patients require clear guidance and continuous support throughout the preoperative, intraoperative, and postoperative phases. This underscores the importance of effective communication, education, and accessible resources for improving patient empowerment and clinical outcomes [ 7 ]. These challenges constrain further enhancement in the precision of diagnosis and treatment, as well as the quality of long-term patient management, creating an urgent need for novel methods and tools to provide assistance. In recent years, the rapid advancement of artificial intelligence (AI) technology, particularly represented by Large Language Models (LLMs), offers a new perspective and potential to address these challenges. In the medical field, LLMs have already demonstrated significant potential in enhancing the quality of medical education, assisting clinical diagnosis, supporting decision-making, and promoting patient health management, among other areas [ 8 ]. Against this backdrop, exploratory applications of LLMs within the specialized field of metabolic and bariatric surgery have begun to emerge. Nevertheless, existing research remains fragmented, lacking comprehensive integration in terms of coverage across different application scenarios, systematic comparison of different model performances, and synthesis of common challenges and future directions faced by the technology. Therefore, this study aims to systematically review the relevant research on the application of LLMs in the field of metabolic and bariatric surgery through a scoping review methodology. Methods Type of review This study was conducted according to the Joanna Briggs Institute methodology for scoping reviews [ 9 ]. Reporting adhered to the preferred reporting items for systematic reviews and meta-analyses extension for scoping reviews (PRISMA-ScR) ( Fig 1 ) [ 10 ]. Fig 1. PRISMA flow chart of the selection process. Open in a new tab Identifying the research question The specific research questions that guided this review were as follows: (i) In which specific scenarios of bariatric surgery are LLMs applied, and how do they perform? (ii) What are the differences in performance among different LLM models when applied in bariatric surgery? (iii) What evidence do existing studies provide regarding the effectiveness of LLMs in bariatric surgery applications? (iv) What are the main challenges currently faced in these applications, and what are the future directions for development? Search strategy A search was conducted in the electronic databases PubMed, Web of Science, The Cochrane Library, Embase, CINAHL, CNKI, Wan Fang, and VIP database, covering literature in both English and Chinese up to November 2025. Common search fields were used, employing a combination of subject headings and free-text keywords. References were also tracked throughout the review process. The full search strategy is provided in Table 1 . Table 1. Search strategy used for each of the databases. PubMed #1 Large Language Models[Mesh] OR Artificial Intelligence[Mesh] #2 (Generative Pre-trained Transformer[tiab] OR ChatGPT[tiab] OR GPT[tiab] OR (Large[tiab] AND Language[tiab] AND Model[tiab]) OR LLM[tiab] OR LLMs[tiab] OR Gemini[tiab] OR DeepSeek[tiab] OR Copilot[tiab] OR LaMDA[tiab] OR LLaMA[tiab] OR BatGPT[tiab] OR Moss[tiab] OR Sora[tiab] OR Kimi[tiab] OR “Ernie Bot”[tiab] OR Qwen*[tiab] OR iFLYTEK Spark[tiab] OR ChatGLM*[tiab] OR Bard[tiab] OR Claude[tiab] OR Bing[tiab] OR Perplexity[tiab] OR mistral[tiab] OR Grok[tiab] OR PaLM[tiab] OR Chatsonic[tiab] OR Jasper[tiab] OR Generative Artificial Intelligence[tiab] OR Generative AI[tiab] OR Chatbot[tiab]) #3 #1 OR #2 #4 Bariatric Surgery[Mesh] OR Obesity, Morbid[Mesh] #5 (Bariatric[tiab] OR Metabolic Surgery[tiab] OR Weight Loss Surgery[tiab] OR Gastric Bypass[tiab] OR Sleeve Gastrectomy[tiab] OR Gastric Sleeve[tiab] OR Gastric Banding[tiab] OR Duodenal Switch[tiab]) #6 #4 OR #5 #7 #3 AND #6 Web of Science #1 TS=(Large Language Models OR Artificial Intelligence OR Generative Pre-trained Transformer OR ChatGPT OR GPT OR LLM OR LLMs OR Gemini OR DeepSeek OR Copilot OR LaMDA OR LLaMA OR BatGPT OR Moss OR Sora OR Kimi OR ErnieBot OR Qwen* OR iFLYTEK Spark OR ChatGLM* OR Bard OR Claude OR Bing OR Perplexity OR mistral OR Grok OR PaLM OR Chatsonic OR Jasper OR Generative Artificial Intelligence OR Generative AI OR Chatbot) #2 TS=(Bariatric Surgery OR Obesity, Morbid OR Bariatric OR Metabolic Surgery OR Weight Loss Surgery OR Gastric Bypass OR Sleeve Gastrectomy OR Gastric Sleeve OR Gastric Banding OR Duodenal Switch) #3 #1 AND #2 Cochrane #1 MeSH descriptor:[Large Language Models] explode all trees #2 MeSH descriptor: [Artificial Intelligence] explode all trees #3 #1 OR #2 #4 (Generative Pre-trained Transformer OR ChatGPT OR GPT OR LLM OR LLMs OR Gemini OR DeepSeek OR Copilot OR LaMDA OR LLaMA OR BatGPT OR Moss OR Sora OR Kimi OR ErnieBot OR Qwen* OR iFLYTEK Spark OR ChatGLM* OR Bard OR Claude OR Bing OR Perplexity OR mistral OR Grok OR PaLM OR Chatsonic OR Jasper OR Generative Artificial Intelligence OR Generative AI OR Chatbot):ti,ab,kw #5 #3 OR #4 #6 MeSH descriptor: [Bariatric Surgery] explode all trees #7 MeSH descriptor: [Obesity, Morbid] explode all trees #8 #6 OR#7 #9 (Bariatric OR Metabolic Surgery OR Weight Loss Surgery OR Gastric Bypass OR Sleeve Gastrectomy OR Gastric Sleeve OR Gastric Banding OR Duodenal Switch):ti,ab,kw #10 #8 OR#9 #11 #5 AND #10 Embase #1 ‘large language model’/exp #2 ‘artificial intelligence’/exp #3 #1 OR #2 #4 generative pre-trained transformer:ti,ab,kw OR chatgpt:ti,ab,kw OR gpt:ti,ab,kw OR llm:ti,ab,kw OR llms:ti,ab,kw OR gemini:ti,ab,kw OR deepseek:ti,ab,kw OR copilot:ti,ab,kw OR lamda:ti,ab,kw OR llama:ti,ab,kw OR batgpt:ti,ab,kw OR moss:ti,ab,kw OR sora:ti,ab,kw OR kimi:ti,ab,kw OR ‘ernie bot’:ti,ab,kw OR qwen*:ti,ab,kw OR ‘iflytek spark’:ti,ab,kw OR chatglm*:ti,ab,kw OR bard:ti,ab,kw OR claude:ti,ab,kw OR bing:ti,ab,kw OR perplexity:ti,ab,kw OR mistral:ti,ab,kw OR grok:ti,ab,kw OR palm:ti,ab,kw OR chatsonic:ti,ab,kw OR jasper:ti,ab,kw OR generative artificial intelligence:ti,ab,kw OR generative ai:ti,ab,kw OR chatbot:ti,ab,kw #5 #3 OR #4 #6 ‘bariatric surgery’/exp #7 ‘morbid obesity’/exp #8 #6 OR #7 #9 bariatric:ti,ab,kw OR metabolic surgery:ti,ab,kw OR weight loss surgery:ti,ab,kw OR gastric bypass:ti,ab,kw OR sleeve gastrectomy:ti,ab,kw OR gastric sleeve:ti,ab,kw OR gastric banding:ti,ab,kw OR duodenal switch:ti,ab,kw #10 #8 OR #9 #11 #5 AND #10 CINAHL S1 MH Large Language Models OR MH Artificial Intelligence S2 TI (“Generative Pre-trained Transformer” OR “ChatGPT” OR “GPT” OR “LLM” OR “LLMs” OR “Gemini” OR “DeepSeek” OR “Copilot” OR “LaMDA” OR “LLaMA” OR “BatGPT” OR “Moss” OR “Sora” OR “Kimi” OR “ErnieBot” OR “Qwen*” OR “iFLYTEK Spark” OR “ChatGLM*” OR “Bard” OR “Claude” OR “Bing” OR “Perplexity” OR “mistral” OR “Grok” OR “PaLM” OR “Chatsonic” OR “Jasper” OR “Generative Artificial Intelligence” OR “Generative AI” OR “Chatbot”) S3 S1 OR S2. S4 MH Bariatric Surgery OR MH Obesity, Morbid S5 TI (“Bariatric” OR “Metabolic Surgery” OR “Weight Loss Surgery” OR “Gastric Bypass” OR “Sleeve Gastrectomy” OR “Gastric Sleeve” OR “Gastric Banding” OR “Duodenal Switch”) S6 S4 OR S5. S7 S3 AND S6. China National Knowledge Infrastructure (CNKI) (Chinese) (Topic: “Large Language Model” OR “Artificial Intelligence” OR “Generative Pre-trained Transformer” OR “GPT” OR “ChatGPT” OR “DeepSeek” OR “Bard” OR “Bing” OR “Copilot” OR “LaMDA” OR “LLaMA” OR “BatGPT” OR “Moss” OR “Sora” OR “Kimi” OR “Generative Artificial Intelligence” OR “Chatbot”) AND (Topic: “Bariatric Surgery” OR “Metabolic Surgery” OR “Obesity Surgery” OR “Gastric Bypass” OR “Sleeve Gastrectomy” OR “Gastric Banding” OR “Duodenal Switch”) WANFANG DATA (Chinese) (Topic: “Large Language Model” OR “Artificial Intelligence” OR “Generative Pre-trained Transformer” OR “GPT” OR “ChatGPT” OR “DeepSeek” OR “Bard” OR “Bing” OR “Copilot” OR “LaMDA” OR “LLaMA” OR “BatGPT” OR “Moss” OR “Sora” OR “Kimi” OR “Generative Artificial Intelligence” OR “Chatbot”) AND (Topic: “Bariatric Surgery” OR “Metabolic Surgery” OR “Obesity Surgery” OR “Gastric Bypass” OR “Sleeve Gastrectomy” OR “Gastric Banding” OR “Duodenal Switch”) VIP database (Chinese) (Topic: “Large Language Model” OR “Artificial Intelligence” OR “Generative Pre-trained Transformer” OR “GPT” OR “ChatGPT” OR “DeepSeek” OR “Bard” OR “Bing” OR “Copilot” OR “LaMDA” OR “LLaMA” OR “BatGPT” OR “Moss” OR “Sora” OR “Kimi” OR “Generative Artificial Intelligence” OR “Chatbot”) AND (Topic: “Bariatric Surgery” OR “Metabolic Surgery” OR “Obesity Surgery” OR “Gastric Bypass” OR “Sleeve Gastrectomy” OR “Gastric Banding” OR “Duodenal Switch”) Open in a new tab Literature inclusion and exclusion criteria Inclusion criteria were determined according to the PCC (population, concept, context) principles [ 11 ]：(i) Participants (P): Involving clinical practice, patient management, or medical education in the field of metabolic and bariatric surgery;(ii) Concept (C): The core of the study involves the application of Large Language Models (LLMs), including but not limited to their development, deployment, evaluation, or comparison. Application forms include answering patient inquiries, generating educational materials, providing clinical decision support, etc.;(iii) Context (C): The application scenario is explicitly limited to metabolic and bariatric surgery. Study types are limited to original research such as quantitative studies, qualitative studies, and mixed-methods studies.Exclusion criteria: (i) Study content not directly related to bariatric surgery or the application of LLMs;(ii) Conference abstracts, commentaries, editorials, systematic reviews, literature reviews, case reports, study protocols, guidelines, or consensus statements;(iii) Literature for which the full text is unavailable, data cannot be extracted, or is non-peer-reviewed;(iv) Literature not published in Chinese or English. Study selection After removing duplicates using EndNote X9 software, literature screening was performed by two researchers, strictly following the inclusion and exclusion criteria. The title and abstract were reviewed first, and the full text of studies potentially meeting the inclusion criteria was further examined. Any disagreements were discussed to reach an agreement, or a third party was consulted. Data extraction The contents were extracted as follows: author, year, country, tool(s), application scenario category, specific application description, and main findings. Results An initial search yielded 1,130 articles. After removing duplicates, screening titles and abstracts, and excluding articles without accessible full texts, 52 articles remained. Following full-text review, 21 English-language articles were ultimately included. The included studies were conducted in multiple countries, including the United States [ 12 – 20 ] (n = 9), China [ 21 , 22 ] (n = 2), Iran [ 23 ] (n = 1), Spain [ 24 , 25 ] (n = 2), Turkey [ 26 , 27 ] (n = 2), Canada [ 28 , 29 ] (n = 2), a multinational collaboration [ 30 ] (n = 1), Germany [ 31 ] (n = 1), and Brazil [ 32 ] (n = 1). The basic characteristics of the included studies are presented in Table 2 . Table 2. Basic characteristics of the literature for inclusion in this analysis (n = 21). Author Year Country Tool(s) Application scenario category Specific application description Main findings Samaan et al. 2023 USA ChatGPT-3.5 Patient Education & Information Model used to answer 151 common bariatric surgery questions to assess accuracy and reproducibility ChatGPT-3.5 showed high accuracy (86.8% “completely accurate”) in answering common questions and good reproducibility (90.7%). Moazzam et al . 2023 USA ChatGPT-3.5 Patient Education & Information Evaluated the quality of model responses to 30 bariatric surgery questions Expert evaluation found ChatGPT-3.5 provided high-quality information; 50% supported integrating it into electronic health systems as a first-line response tool. Aburumman et al. 2024 USA ChatGPT-3.5 Patient Education & Information Compared model responses with hospital website answers to 8 ESG-related questions Experts accurately identified AI-generated answers 54% of the time; overall, no significant differences were found between AI and website answers across assessed dimensions. Srinivasan et al. 2024 USA GPT-3.5, GPT-4, Bard Patient Education & Information Compared readability of bariatric surgery materials generated and simplified by three models GPT-4 effectively simplified professional texts to a 6th-9th grade reading level with minimal loss of accuracy; Bard’s simplification sometimes reduced answer completeness. Samaan et al. 2024 USA GPT-3.5, GPT-4 Patient Education & Information Compared accuracy and comprehensiveness of the two models in answering 151 bariatric surgery questions GPT-4 matched GPT-3.5’s accuracy while providing more comprehensive information in 36.4% of responses. Sanders et al. 2024 USA ChatGPT-4 Knowledge Assessment & Exams Tested model performance on metabolic and bariatric surgery simulated board examination questions ChatGPT-4 achieved a 74.1% correct answer rate on specialty simulation questions, showing stable performance across different knowledge categories. Ozmen et al. 2025 USA BariatricSurgeryGPT (fine-tuned GPT-2) Domain-Specific Model Development Fine-tuned model with professional abstracts; evaluated its ability to generate specialty-specific content The fine-tuned domain-specific model significantly outperformed the general GPT-2 model across all metrics for generating professional text, showing higher semantic relevance, terminology accuracy, and information completeness. Annor et al. 2025 USA Gemini, Copilot Patient Education & Information Evaluated the appropriateness and completeness of two models’ responses to weight management guideline-related questions Copilot provided more complete answers but exhibited guideline bias; Gemini was more cautious, sometimes refusing to answer or providing incomplete information. Kahlon et al. 2025 USA ChatGPT-4 Clinical Decision Support Compared model treatment recommendations with physician recommendations in five different bariatric surgery clinical scenarios ChatGPT-4 showed good agreement with physician recommendations in simple clinical scenarios, but significant discrepancies were noted in complex cases. Leng et al. 2025 China ChatGPT-4 Patient Education & Information Designed 30 questions covering six domains of bariatric surgery; assessed the model’s feasibility as a knowledge resource ChatGPT-4 performed well in bariatric surgery Q&A, with 50% of responses receiving the highest score, but limitations included outdated criteria, lack of specificity, and suboptimal response structure. Guo et al. 2025 China ChatGPT-4, DeepSeek Patient Education & Information Compared the quality, reliability, and readability of patient education materials generated by two models ChatGPT significantly outperformed DeepSeek in material quality and reliability; however, the readability of both models’ outputs did not reach patient-friendly levels. Mahjoubi et al. 2025 Iran ChatGPT-4, DALL·E 3 Medical Image Processing Evaluated model accuracy in identifying surgical procedures and generating surgical illustrations ChatGPT-4 and DALL·E 3 demonstrated low accuracy in identifying and generating bariatric surgery images, with issues such as missing anastomoses, incorrect anatomical depictions, and terminology errors in textual descriptions. Lopez-Gonzalez et al. 2024 Spain ChatGPT-4 Clinical Decision Support Compared model-recommended surgical techniques with hospital algorithm decisions for 161 patients undergoing MBS The match rate between ChatGPT-4 recommendations and the hospital algorithm decisions was only 34.16%, showing no significant correlation. Sanchez-Cordero et al. 2025 Spain ChatGPT Clinical Decision Support Evaluated consistency between model recommendations and clinical decisions for 283 primary bariatric surgery patients, pre- and post-exposure to scientific literature Even after literature-based training, ChatGPT’s recommendations showed low agreement with clinical decisions (25.8%). Its outputs aligned more closely with global surgical trends, while individual patient matching remained limited. Aksoy et al. 2025 Turkey ChatGPT-4.0, ChatGPT-Omni, Gemini Patient Education & Information Evaluated three models’ performance on 720 One Anastomosis Gastric Bypass (OAGB) related clinical questions ChatGPT-Omni demonstrated the best performance in answering OAGB-related clinical questions, with higher precision, recall, and F1-scores compared to the other models. Kumru Yildirim et al. 2025 Turkey ChatGPT-3.5 Patient Education & Information Evaluated consistency between model-provided postoperative nutrition advice for 27 questions and clinical guidelines ChatGPT-3.5’s postoperative nutrition advice was fully consistent with clinical guidelines in only 44.4% of cases; it showed weak personalization capability and made typical errors, such as misclassifying leafy greens as “high-protein food.” Lee et al. 2024 Canada ChatGPT-4, Bing, Bard Patient Education & Information Evaluated appropriateness and readability of three models’ responses to 36 clinical guideline-related questions ChatGPT-4 had a significantly higher proportion of appropriate responses (85.7%) compared to Bard (74.3%) and Bing (25.7%), yet all remained in the “fairly difficult to read” range. ChatGPT-4 performed best in categories like surgical techniques and perioperative management. Lee et al. 2024 Canada ChatGPT-4, Bing, Bard Knowledge Assessment & Exams Tested the accuracy of three models on 200 multiple-choice questions in bariatric surgery ChatGPT-4 achieved the highest overall accuracy (83.0%), followed by Bard (76.0%) and Bing (65.5%). ChatGPT-4 also had the highest correct rates in “Treatment & Procedures” (83.1%) and “Complications” (91.7%) categories, and left no questions unanswered. Jazi et al. 2023 Multinational ChatGPT-4 Clinical Decision Support Compared model surgical recommendations with multinational expert consensus in 10 clinical cases Model recommendations aligned with majority expert opinion in only 30% of cases; in 40% of cases, the model gave inconsistent answers across two attempts (reproducibility 60%). Vedder et al. 2025 Germany GPT-4o Patient Education & Information Compared model vs. human expert answers to 200 real patient questions on accuracy, completeness, empathy, and patient satisfaction Model response time (2.7 seconds) was significantly shorter than human experts (87.2 seconds), and answer length (607 characters) was significantly longer than human experts (262 characters). Patients rated model answers significantly higher for clarity, completeness, and empathy; 64.9% preferred model responses. Bigolin et al. 2025 Brazil ChatGPT-3.5, ChatGPT-4 Patient Education & Information Evaluated quality, empathy, safety, and clinical acceptability of two model versions answering 10 postoperative questions Both versions scored >7 on average across all questions. ChatGPT-4 significantly outperformed v3.5 in “satisfaction” and “acceptability” (P < 0.05). Both versions demonstrated good empathetic capabilities. Open in a new tab Application scenarios of LLMs in bariatric surgery This review included a total of 21 studies. The application of Large Language Models in bariatric surgery primarily focused on the following scenarios: thirteen studies [ 12 – 16 , 19 , 21 , 22 , 26 – 28 , 31 , 32 ] concentrated on the patient education and information consultation scenario. This mainly involved providing patients with knowledge Q&A related to bariatric surgery, generating patient education materials, and optimizing text readability. This scenario represents the most widely explored area of LLM application, particularly demonstrating potential in improving the accessibility of disease-related information and patient engagement. Four studies [ 20 , 24 , 25 , 30 ] explored their application in clinical decision support, primarily focusing on recommendations for surgical techniques. However, consistency between LLMs and clinical guidelines or expert consensus in such tasks was low, with particularly limited performance in complex cases. Two studies [ 17 , 29 ] discussed the application of LLMs in bariatric surgery professional knowledge assessment and examinations, using them to simulate specialty board exam questions and evaluate the grasp of medical knowledge. Research indicated that LLMs performed well on standardized test questions but remained insufficient in clinical reasoning questions. One study [ 23 ] focused on medical image processing, attempting to use LLMs for bariatric surgery image recognition and generation. However, the results showed low anatomical accuracy, indicating that they are not yet suitable for clinical or educational purposes. One study [ 18 ] explored the development of a domain-specific Large Language Model for bariatric surgery. The specialty model, built by fine-tuning a general-purpose model, outperformed the base model in professional text generation tasks, demonstrating the potential of vertical domain optimization. Performance differences among different LLMs The performance of different LLMs in bariatric surgery tasks showed significant variation. Among the ChatGPT series models, GPT-4 demonstrated higher informational accuracy in most studies and outperformed its predecessor GPT-3.5 in providing detailed, contextualized responses [ 16 ]. However, this series of models commonly exhibited issues such as initially low readability [ 15 ]and knowledge update lag [ 21 ]. The performance of other general-purpose models like Gemini, Bard, and DeepSeek varied: Gemini demonstrated higher caution in some studies, sometimes refusing to answer sensitive medical questions directly, though this could also lead to insufficient completeness in its responses [ 19 ], DeepSeek slightly outperformed ChatGPT in text readability but lagged significantly behind in information quality and reliability [ 22 ], while Bard and Bing models showed unstable performance across multiple studies [ 28 , 29 ], particularly displaying low appropriateness in clinical advice generation tasks [ 28 ]. Furthermore, domain-fine-tuned models such as BariatricSurgery GPT excelled in the accuracy of professional terminology and semantic relevance [ 18 ], highlighting the potential of enhancing model specialization through vertical domain optimization. Overall, the differences in model performance depend not only on their underlying architecture and training data but are also closely related to the task type, prompt engineering, and evaluation criteria. Evaluation of LLM application effectiveness in bariatric surgery In terms of accuracy, LLMs performed well in common knowledge Q&A tasks. For instance, the ChatGPT series achieved an accuracy rate of up to 86.8% on common bariatric surgery questions [ 12 ]. However, in clinical decision support tasks requiring clinical judgment, their agreement rates with expert opinions or real clinical decisions were typically below 40% [ 24 , 25 , 30 ], with particularly low matching rates for personalized surgical recommendations [ 20 , 24 ]. Readability remains a significant concern. Studies have pointed out that the average reading level of LLM-generated texts often exceeds the 6th to 8th-grade level recommended by the American Medical Association, mostly falling between the 9th-grade and college levels [ 15 , 22 ]. Although targeted prompts can simplify texts to some extent, bringing the readability level down to grades 6–9 [ 15 ], consistently meeting patient-friendly reading standards remains challenging, and some responses suffer from structural verbosity and excessive detail [ 21 ]. Regarding empathy and patient satisfaction, a few studies have conducted evaluations. One study found that GPT-4o received significantly higher patient ratings for answer clarity, completeness, and empathy, with 64.9% of patients expressing a preference for AI-generated responses [ 31 ]. Another study also noted that ChatGPT scored high on empathy, and patient satisfaction and acceptability of AI responses were good [ 32 ]. In terms of information quality, ChatGPT’s response quality generally surpassed that of other models [ 13 , 22 ], but issues such as missing citations and lack of transparency regarding information sources remain prevalent. Domain-specific models demonstrated superior performance in the semantic relevance and professionalism of generated content [ 18 ]. In terms of efficiency, LLMs demonstrated a clear advantage, with an average response generation time of only a few seconds, significantly shorter than the time required for clinicians to draft similar content [ 31 ], suggesting their potential value in improving clinical workflow efficiency. Discussion Performance variation and capability boundaries across application scenarios Current research indicates a clear divergence in the efficacy of LLMs across different application scenarios within bariatric surgery. In the realm of patient education and information consultation, LLMs, leveraging their robust natural language generation capabilities, can provide patients with comprehensive and accurate knowledge-related responses about surgery. Multiple studies demonstrate that their response quality has reached a relatively high level, with some models achieving accuracy rates exceeding 85% in relevant evaluations [ 12 ]. This aligns with the findings of Goudrar et al.[ 33 ], suggesting that LLMs can serve as a beneficial supplementary resource for patients seeking information on bariatric surgery and possess the potential to become auxiliary clinical education tools. However, in scenarios requiring clinical judgment, particularly for tasks involving personalized surgical plan recommendations and decision support, the performance of existing models is significantly constrained. Their alignment with clinical guidelines or expert consensus is generally low. Lopez-Gonzalez et al.[ 24 ] reported only 34.16% concordance between GPT-4 and hospital algorithm decisions. Several factors contributed: ChatGPT-4’s knowledge is updated only until April 2023, it can only access open-access articles, and it explicitly states it is not designed for professional medical use. Sanchez-Cordero et al.[ 25 ] found that even after contextual training with 412 scientific articles, concordance improved only from 20.0% to 25.8%. Notably, ChatGPT tended to mirror the global “average” procedure distribution rather than tailoring recommendations to individual patients. The authors also noted that using a single center’s practice as the gold standard may introduce bias, as surgical choices vary across centers. Kahlon et al.[ 20 ] observed only fair agreement between ChatGPT-4 and bariatric surgeons, with moderate inconsistency across two runs. Surgeons can integrate nuanced, patient-specific information that AI cannot fully weigh, highlighting a key limitation. Jazi et al.[ 30 ] reported that ChatGPT-4 matched expert consensus in only 30% of complex cases, and gave inconsistent answers in 40% of scenarios. The model failed to recognize critical patient-specific risk factors. This issue extends beyond bariatric surgery. Research by de Menezes Torres et al.[ 34 ] on the use of the large language model ChatGPT in oral and maxillofacial surgery highlights its limitations in handling complex clinical decisions and providing personalized recommendations for cases such as oral cancer and orthognathic surgery. In summary, current LLMs lack individualized, context-aware reasoning; exhibit output instability; have outdated knowledge bases and limited access to full evidence; and cannot handle multi-criteria trade-offs. This underscores that while LLMs may be useful as educational tools, they are currently not reliable for autonomous surgical decision-making, especially in complex clinical contexts, highlighting the value of human expertise [ 26 ]. This performance variation clearly delineates the current capability boundaries of LLMs: they excel at processing and generating structured medical knowledge but currently lack a comprehensive and personalized perspective for tasks requiring the integration of multi-dimensional information, understanding complex clinical contexts, and performing dynamic reasoning, indicating fundamental limitations remain. Potential value and integration positioning in clinical practice LLMs demonstrate multifaceted supportive potential in bariatric surgery. Firstly, they can significantly enhance clinical workflow efficiency, for instance, by automating the generation of patient education materials and rapidly answering common consultation questions, thereby freeing healthcare professionals from repetitive informational tasks. By improving answer efficiency under physician supervision while maintaining accuracy, LLMs can optimize doctors’ time management and enhance patient satisfaction in bariatric care communication [ 31 ]. They can serve as supplementary resources for patient education, clinician assistance, and public health promotion [ 21 ]. Secondly, they facilitate the dissemination of health information. Particularly in contexts of unevenly distributed medical resources, they can provide patients with timely and reliable basic medical knowledge, helping to bridge service gaps caused by geographical or economic disparities. A study on LLM applications in ophthalmology showed that large language models like GPT-4 could assist ophthalmologists in distinguishing urgent from routine visits, improving remote ophthalmic triage in low-resource settings [ 35 ], thereby offering remote services to patients. More importantly, LLMs performed excellently on metabolic and bariatric surgery specialty certification simulation questions, achieving a correct answer rate of 74.1% with consistent performance across different knowledge categories [ 17 ], indicating AI’s potential application value in specialty medical education and exam preparation. They can serve as effective adjuncts in medical education, aiding in the training of medical students and residents. It is crucial to clarify that at this stage, LLMs should be positioned as “assistive tools” rather than “decision-making agents.” Their core value lies in augmenting, not replacing, the professional judgment of clinical experts. Challenges and future directions The foremost issues are insufficient information accuracy and timeliness. The trustworthiness and safety of their outputs remain inadequate, posing risks such as hallucination generation, outdated information, inconsistency with the latest guidelines, low readability, and potential biases. This aligns with the findings of Sanker et al.[ 36 ]. Secondly, LLMs lack a deep understanding of complex clinical contexts and struggle to handle patient cases involving individualized factors like multiple comorbidities or previous surgical history, limiting their utility in clinical decision support. Furthermore, model outputs often suffer from poor interpretability; the reasoning process behind their generated conclusions is opaque, making it difficult for clinicians to trace and verify the information basis. These technical limitations are closely intertwined with broader ethical and safety considerations, which deserve detailed elaboration. First, data privacy and confidentiality are critical concerns, as noted in multiple studies; patient queries often contain sensitive health information, and inputting such data into public or unencrypted LLM interfaces risks unauthorized access or data breaches, especially because some models store conversations for retraining, and the lack of transparency regarding how platforms handle personal data further exacerbates these risks. Second, algorithmic bias may lead to unfair or inaccurate recommendations; LLMs trained on non‑representative datasets may generate inappropriate recommendations, thereby exacerbating the risk of healthcare disparities. Third, liability and accountability remain unclear; the “black‑box” nature of LLMs makes their reasoning process opaque, and when an LLM contributes to a clinical decision that results in an adverse event, responsibility is ambiguous—it is unclear whether the developer, the hospital, the supervising physician, or the user should be held accountable, as current legal frameworks lack clear guidance for LLM‑assisted surgical decisions. Fourth, over‑reliance on LLMs may negatively affect the patient‑clinician relationship, as patients might reduce trust in human providers or delay necessary consultations. Finally, the most frequently cited issue is the “hallucination” phenomenon—LLMs tend to generate responses with high confidence and coherent structure even when the content is factually incorrect, which may mislead patients and lead uninformed users to adopt erroneous and potentially dangerous advice. Taken together, these risks emphasize that LLMs are not yet mature and must currently be positioned as supportive tools under human supervision, not autonomous advisors. Future development directions should include: integrating AI with human healthcare professionals as a key pathway to optimizing patient care [ 37 ], establishing evaluation standards oriented towards real-world clinical scenarios, exploring safe and effective human-AI collaborative workflows, and constructing comprehensive ethical governance frameworks. Specifically, future research should focus on three interconnected priorities. First, performance validation and multidimensional evaluation. Given the low consistency of LLMs in complex clinical decisions and their tendency to generate hallucinations or outdated information, rigorous and continuous evaluation and improvement are needed, including testing different models and prompt strategies. Special attention should be paid to reducing hallucinations, improving reproducibility, and enabling individualized recommendations. Assessment frameworks must systematically incorporate core metrics such as accuracy, reproducibility, readability, bias detection, and clinical utility, with standardized evaluation protocols. Second, development of human‑AI collaboration models. Research should use various data sources to explore optimal integration patterns that balance efficiency and safety. Different collaboration models should be tested continuously across diverse simulated clinical scenarios, evaluating their impact on decision quality, workflow efficiency, clinician acceptance, and patient outcomes. Third, establishment of ethical and legal governance frameworks. Clear operational guidelines are urgently needed: for data privacy, secure handling of patient queries and preventing unauthorized use of data for other purposes; for liability allocation, defining responsible parties when LLM suggestions contribute to adverse events and establishing a clear accountability framework; and for informed consent, disclosing AI use to patients. Only through the synergistic advancement of technological innovation, clinical validation, and regulatory management can the responsible application and sustainable development of LLMs in this field be achieved. Conclusions Large Language Models have demonstrated potential as auxiliary informational tools and communication mediators in bariatric surgery, showing particular value in enhancing information accessibility, supporting doctor-patient communication, and medical education. However, their current capabilities remain confined to structured knowledge transmission and are not yet reliable for clinical tasks requiring professional judgment, personalized decision-making, and comprehension of complex contexts. Future development should focus on the specialized optimization of models, standardization of evaluation systems, and exploration of clinical integration pathways. This will facilitate the deep integration of AI with bariatric surgery under the premise of ensuring safety, reliability, and equity. Supporting information S1 Checklist. Preferred reporting items for systematic reviews and meta-analyses extension for scoping reviews (PRISMA-ScR) checklist. (DOCX) pone.0350748.s001.docx (769.9KB, docx) Data Availability All relevant data are within the manuscript and its Supporting Information files. Funding Statement The author(s) received no specific funding for this work. References 1. CDC. Adult obesity facts. Obesity. Accessed 2026 February 11. https://www.cdc.gov/obesity/adult-obesity-facts/index.html 2. Eisenberg D, Shikora SA, Aarts E, Aminian A, Angrisani L, Cohen RV, et al. 2022 American society of metabolic and bariatric surgery (ASMBS) and international federation for the surgery of obesity and metabolic disorders (IFSO) indications for metabolic and bariatric surgery. Obes Surg. 2023;33(1):3–14. doi: 10.1007/s11695-022-06332-1 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 3. Arterburn DE, Telem DA, Kushner RF, Courcoulas AP. Benefits and risks of bariatric surgery in adults: a review. JAMA. 2020;324(9):879–87. doi: 10.1001/jama.2020.12567 [ DOI ] [ PubMed ] [ Google Scholar ] 4. Schauer PR, Bhatt DL, Kirwan JP, Wolski K, Aminian A, Brethauer SA, et al. Bariatric surgery versus intensive medical therapy for diabetes - 5-year outcomes. N Engl J Med. 2017;376(7):641–51. doi: 10.1056/NEJMoa1600869 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 5. Chinese Society for Metabolic and Bariatric Surgery (CSMBS), Chinese Society for Integrated Health of Metabolic and Bariatric Surgery (CSMBS IH), Chinese Obesity and Metabolic Surgery Collaborative (COMES Collaborative). Chinese obesity and metabolic surgery database: annual report 2023. 2024. 10.3877/cma.j.issn.2095-9605.2024.02.001 [ DOI ] 6. Topart P. Obesity surgery: Which procedure should we choose and why?. J Visc Surg. 2023;160(2S):S30–7. doi: 10.1016/j.jviscsurg.2022.12.010 [ DOI ] [ PubMed ] [ Google Scholar ] 7. Scarano Pereira JP, Martinino A, Manicone F, Scarano Pereira ML, Iglesias Puzas Á, Pouwels S, et al. Bariatric surgery on social media: A cross-sectional study. Obes Res Clin Pract. 2022;16(2):158–62. doi: 10.1016/j.orcp.2022.02.005 [ DOI ] [ PubMed ] [ Google Scholar ] 8. Jeblick K, Schachtner B, Dexl J, Mittermeier A, Stüber AT, Topalis J, et al. ChatGPT makes medicine easy to swallow: an exploratory case study on simplified radiology reports. Eur Radiol. 2024;34(5):2817–25. doi: 10.1007/s00330-023-10213-1 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 9. Peters MDJ, Marnie C, Tricco AC, Pollock D, Munn Z, Alexander L, et al. Updated methodological guidance for the conduct of scoping reviews. JBI Evid Synth. 2020;18(10):2119–26. doi: 10.11124/JBIES-20-00167 [ DOI ] [ PubMed ] [ Google Scholar ] 10. Tricco AC, Lillie E, Zarin W, O’Brien KK, Colquhoun H, Levac D, et al. PRISMA extension for scoping reviews (PRISMA-ScR): checklist and explanation. Ann Intern Med. 2018;169(7):467–73. doi: 10.7326/M18-0850 [ DOI ] [ PubMed ] [ Google Scholar ] 11. Lockwood C, Dos Santos KB, Pap R. Practical guidance for knowledge synthesis: scoping review methods. Asian Nurs Res (Korean Soc Nurs Sci). 2019;13(5):287–94. doi: 10.1016/j.anr.2019.11.002 [ DOI ] [ PubMed ] [ Google Scholar ] 12. Samaan JS, Yeo YH, Rajeev N, Hawley L, Abel S, Ng WH, et al. Assessing the accuracy of responses by the language model ChatGPT to questions regarding bariatric surgery. Obes Surg. 2023;33(6):1790–6. doi: 10.1007/s11695-023-06603-5 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 13. Moazzam Z, Lima HA, Endo Y, Noria S, Needleman B, Pawlik TM. A paradigm shift: online artificial intelligence platforms as an informational resource in bariatric surgery. Obes Surg. 2023;33(8):2611–4. doi: 10.1007/s11695-023-06675-3 [ DOI ] [ PubMed ] [ Google Scholar ] 14. Aburumman R, Al Annan K, Mrad R, Brunaldi VO, Gala K, Abu Dayyeh BK. Assessing ChatGPT vs. standard medical resources for endoscopic sleeve gastroplasty education: a medical professional evaluation study. Obes Surg. 2024;34(7):2718–24. doi: 10.1007/s11695-024-07283-5 [ DOI ] [ PubMed ] [ Google Scholar ] 15. Srinivasan N, Samaan JS, Rajeev ND, Kanu MU, Yeo YH, Samakar K. Large language models and bariatric surgery patient education: a comparative readability analysis of GPT-3.5, GPT-4, Bard, and online institutional resources. Surg Endosc. 2024;38(5):2522–32. doi: 10.1007/s00464-024-10720-2 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 16. Samaan JS, Rajeev N, Ng WH, Srinivasan N, Busam JA, Yeo YH, et al. ChatGPT as a source of information for bariatric surgery patients: a comparative analysis of accuracy and comprehensiveness between GPT-4 and GPT-3.5. Obes Surg. 2024;34(5):1987–9. doi: 10.1007/s11695-024-07212-6 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 17. Sanders A, Lim R, Jones D, Vosburg RW. Artificial intelligence large language model scores highly on focused practice designation in metabolic and bariatric surgery board practice questions. Surg Endosc. 2024;38(11):6678–81. doi: 10.1007/s00464-024-11267-y [ DOI ] [ PubMed ] [ Google Scholar ] 18. Ozmen BB, Berber I, Dang JT, Schwarz GS, Kroh M. Development of a bariatric surgery specific artificial intelligence large language model: bariatricsurgeryGPT. Surg Innov. 2026;33(3):276–82. doi: 10.1177/15533506251400130 [ DOI ] [ PubMed ] [ Google Scholar ] 19. Annor E, Atarere J, Ubah N, Jolaoye O, Kunkle B, Egbo O, et al. Assessing online chat-based artificial intelligence models for weight loss recommendation appropriateness and bias in the presence of guideline incongruence. Int J Obes (Lond). 2025;49(5):896–901. doi: 10.1038/s41366-025-01717-5 [ DOI ] [ PubMed ] [ Google Scholar ] 20. Kahlon S, Sleet M, Sujka J, Docimo S, DuCoin C, Dimou F, et al. Evaluating the concordance of ChatGPT and physician recommendations for bariatric surgery. Can J Physiol Pharmacol. 2025;103(2):70–4. doi: 10.1139/cjpp-2024-0026 [ DOI ] [ PubMed ] [ Google Scholar ] 21. Leng Y, Yang Y, Liu J, Jiang J, Zhou C. Evaluating the Feasibility of ChatGPT-4 as a Knowledge Resource in Bariatric Surgery: A Preliminary Assessment. Obes Surg. 2025;35(2):645–50. doi: 10.1007/s11695-024-07666-8 [ DOI ] [ PubMed ] [ Google Scholar ] 22. Guo S, Yang C-L, Lin X-P, Jiang M, Chen J, Tuo K-X, et al. Evaluating artificial intelligence-generated patient education materials for bariatric surgery: comparative analysis of response quality, reliability, and readability across ChatGPT and deepseek models. Obes Surg. 2025;35(11):4628–38. doi: 10.1007/s11695-025-08249-x [ DOI ] [ PubMed ] [ Google Scholar ] 23. Mahjoubi M, Shahabi S, Sheikhbahaei S, Jazi AHD. Evaluating AI capabilities in bariatric surgery: a study on ChatGPT-4 and DALL·E 3’s recognition and illustration accuracy. Obes Surg. 2025;35(2):638–41. doi: 10.1007/s11695-024-07653-z [ DOI ] [ PubMed ] [ Google Scholar ] 24. Lopez-Gonzalez R, Sanchez-Cordero S, Pujol-Gebellí J, Castellvi J. Evaluation of the impact of ChatGPT on the selection of surgical technique in bariatric surgery. Obes Surg. 2025;35(1):19–24. doi: 10.1007/s11695-024-07279-1 [ DOI ] [ PubMed ] [ Google Scholar ] 25. Sanchez-Cordero S, Lopez-Gonzalez R, Fernandez H, Pujol-Gebellí J. Training ChatGPT for surgical decisions: Bariatric surgery analysis using algorithms and evidence. Obes Res Clin Pract. 2025;19(4):352–5. doi: 10.1016/j.orcp.2025.08.002 [ DOI ] [ PubMed ] [ Google Scholar ] 26. Aksoy E. The Performance of artificial intelligence in one anastomosis gastric bypass surgery: comparative efficacy of ChatGPT-4.0, ChatGPT-Omni, and Gemini AI. Obes Surg. 2025;35(4):1469–75. doi: 10.1007/s11695-025-07794-9 [ DOI ] [ PubMed ] [ Google Scholar ] 27. Kumru Yildirim M, Bayam H, Yildiz N, Kahraman Gök F, Batar N. Evaluation of nutritional recommendations provided by the ChatGPT language model to bariatric surgery patients. Rom J Intern Med. 2025;63(2):185–7. doi: 10.2478/rjim-2025-0004 [ DOI ] [ PubMed ] [ Google Scholar ] 28. Lee Y, Shin T, Tessier L, Javidan A, Jung J, Hong D, et al. Harnessing artificial intelligence in bariatric surgery: comparative analysis of ChatGPT-4, Bing, and Bard in generating clinician-level bariatric surgery recommendations. Surg Obes Relat Dis. 2024;20(7):603–8. doi: 10.1016/j.soard.2024.03.011 [ DOI ] [ PubMed ] [ Google Scholar ] 29. Lee Y, Tessier L, Brar K, Malone S, Jin D, McKechnie T, et al. Performance of artificial intelligence in bariatric surgery: comparative analysis of ChatGPT-4, Bing, and Bard in the American Society for Metabolic and Bariatric Surgery textbook of bariatric surgery questions. Surg Obes Relat Dis. 2024;20(7):609–13. doi: 10.1016/j.soard.2024.04.014 [ DOI ] [ PubMed ] [ Google Scholar ] 30. Jazi AHD, Mahjoubi M, Shahabi S, Alqahtani AR, Haddad A, Pazouki A, et al. Bariatric evaluation through AI: a survey of expert opinions versus ChatGPT-4 (BETA-SEOV). Obes Surg. 2023;33(12):3971–80. doi: 10.1007/s11695-023-06903-w [ DOI ] [ PubMed ] [ Google Scholar ] 31. Vedder K, Blank S, Wilhelm T, Fidan D, Pachkiv I, Cao H, et al. Comparative analysis of large language model and physician-generated responses in bariatric patient inquiries: assessing the accuracy and patient satisfaction. Obes Surg. 2025;35(9):3801–9. doi: 10.1007/s11695-025-08115-w [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 32. Bigolin AV, Nunes Chibiaque de Lima J, Machado Grossi JV, Hartmann Rost I, Machado da Rosa M, Piccolotto Concolatto F. Could ChatGPT Be a Tool Capable of Providing Qualified, Empathetic, and Assertive Answers to Patients After Bariatric Surgery? A Comparative Analysis of Its Versions. Obes Surg. 2025;35(11):4605–11. doi: 10.1007/s11695-025-08306-5 [ DOI ] [ PubMed ] [ Google Scholar ] 33. Goudrar R, Zekraoui O, Moussa I, Nguyen D-D, Bouhadana D, Li T, et al. Large language model chatbots for patient education in kidney stones: a scoping review. World J Urol. 2025;43(1):641. doi: 10.1007/s00345-025-06019-z [ DOI ] [ PubMed ] [ Google Scholar ] 34. de Menezes Torres LM, de Morais EF, Fernandes Almeida DR de M, Pagotto LEC, de Santana Santos T. The impact of the large language model ChatGPT in oral and maxillofacial surgery: a systematic review. Br J Oral Maxillofac Surg. 2025;63(5):357–62. doi: 10.1016/j.bjoms.2025.03.006 [ DOI ] [ PubMed ] [ Google Scholar ] 35. Cohen L, Gupta AR, Patel P, Gill GS, Bains H, Gupta S. The Role of Large Language Models in Ophthalmology: A Review of Current Applications, Performance, and Future Directions. Cureus. 2025;17(11):e97374. doi: 10.7759/cureus.97374 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 36. Sanker V, Nordin EOR, Heesen P, Elfadali MA, Anwar M, Chintapalli RD, et al. Current trends and future prospects of language models and processing systems in spine surgery - a scoping review. Neurosurg Rev. 2025;48(1):633. doi: 10.1007/s10143-025-03785-7 [ DOI ] [ PubMed ] [ Google Scholar ] 37. Dahiya DS, Ali H, Moond V, Shah MDA, Santana C, Ali N, et al. Large language models in gastroenterology and gastrointestinal surgery: a new frontier in patient communication and education. Gastroenterology Res. 2025;18(2):39–48. doi: 10.14740/gr2011 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] PLoS One. doi: 10.1371/journal.pone.0350748.r001 Decision Letter 0 Hongyang Ma Author information Copyright and License information Roles Hongyang Ma : Guest Editor © 2026 Hongyang Ma This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. PMC Copyright notice 27 Apr 2026 -->PONE-D-26-11116-->-->The Application of Large Language Models in Bariatric Surgery: A Scoping Review-->-->PLOS One Dear Dr. Zheng, Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process. Please submit your revised manuscript by Jun 11 2026 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org . When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file. Please include the following items when submitting your revised manuscript:--> A letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'. A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'. An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'. --> If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter. If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols . As the corresponding author, your ORCID iD is verified in the submission system and will appear in the published article. PLOS supports the use of ORCID, and we encourage all coauthors to register for an ORCID iD and use it as well. Please encourage your coauthors to verify their ORCID iD within the submission system before final acceptance, as unverified ORCID iDs will not appear in the published article. Only the individual author can complete the verification step; PLOS staff cannot verify ORCID iDs on behalf of authors. We look forward to receiving your revised manuscript. Kind regards, Hongyang Ma, Ph.D., D.D.S Guest Editor PLOS One Journal Requirements: When submitting your revision, we need you to address these additional requirements. 1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf 2. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information . 3. If the reviewer comments include a recommendation to cite specific previously published works, please review and evaluate these publications to determine whether they are relevant and should be cited. There is no requirement to cite these works unless the editor has indicated otherwise. 4. Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice. [Note: HTML markup is below. Please do not edit.] Reviewers' comments: Reviewer's Responses to Questions --> Comments to the Author 1. Is the manuscript technically sound, and do the data support the conclusions? The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented. --> Reviewer #1: Yes ********** -->2. Has the statistical analysis been performed appropriately and rigorously? --> Reviewer #1: Yes ********** -->3. Have the authors made all data underlying the findings in their manuscript fully available? The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.--> Reviewer #1: Yes ********** -->4. Is the manuscript presented in an intelligible fashion and written in standard English? PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.--> Reviewer #1: Yes ********** -->5. Review Comments to the Author Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)--> Reviewer #1: This article is a comprehensive scoping review supported by an extensive literature review. Analyzing 21 English-language studies, it reveals the applications of LLMs in areas such as patient education, information counseling, clinical decision support systems, and professional knowledge assessment. The study fills a significant gap in the field by exploring the potential use of LLMs in surgical applications. The scoping review methodology ensures a comprehensive approach to the subject, systematically presenting key findings and applications from the existing literature. However, the following points require improvement: 1. The consistency of LLMs in clinical decision-making processes was found to be low. This issue needs to be discussed in more detail. 2. The section on recommendations for future research could be more comprehensive. Specifically, areas requiring further study should be identified. 3. Providing more detail on the ethical and safety risks associated with the use of LLMs would be beneficial in addressing concerns in this area. The article presents important findings and contributes to the field. Correcting the weaknesses mentioned above will improve the publishability of the article. It is important for the authors to review the work taking the suggested corrections into consideration. The article is considered suitable for publication after the necessary corrections have been made. ********** -->6. PLOS authors have the option to publish the peer review history of their article ( what does this mean? ). If published, this will include your full peer review and any attached files. If you choose “no”, your identity will remain anonymous but your review may still be made public. Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy .--> Reviewer #1: Yes: Mustafa Cem Algin ********** [NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.] To ensure your figures meet our technical requirements, please review our figure guidelines: https://journals.plos.org/plosone/s/figures You may also use PLOS’s free figure tool, NAAS, to help you prepare publication quality figures: https://journals.plos.org/plosone/s/figures#loc-tools-for-figure-preparation . NAAS will assess whether your figures meet our technical requirements by comparing each figure against our figure specifications. -->-->--> -->--> PLoS One. 2026 Jun 5;21(6):e0350748. doi: 10.1371/journal.pone.0350748.r002 Author response to Decision Letter 1 Article notes Copyright and License information Collection date 2026. PMC Copyright notice 8 May 2026 Dear Editor and Reviewers, We sincerely appreciate the time and expertise invested in reviewing our manuscript 【PONE-D-26-11116. Thank you for your constructive feedback, which has significantly strengthened the quality of this work. We have carefully addressed all comments and revised the manuscript accordingly. Below, we provide a point-by-point response to each suggestion, with corresponding revisions highlighted in the tracked-changes version of the manuscript. Journal Requirements: 1.“Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. ” Response: We have revised the manuscript to fully comply with PLOS ONE’s formatting guidelines. Should any formatting inconsistencies persist despite our rigorous checks using the PLOS ONE templates, we remain fully open to implementing additional corrections as directed by the editorial office. A 24-hour revision turnaround is guaranteed upon request. 2.“Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly.” Response: 【line 523】 ①　A dedicated paragraph titled "Supporting Information " has been inserted after the References section. ②　Captions for all Supporting Information files (e.g., S1 Table) have been added under the "Supporting Information" section at the end of the manuscript. ③　In-text citations to Supporting Information have been standardized 3.If the reviewer comments include a recommendation to cite specific previouslypublished works, please review and evaluate these publications to determine whetherthey are relevant and should be cited. There is no requirement to cite these works unless the editor has indicated otherwise. We thank the editor for this reminder. We have carefully reviewed all comments from the reviewer and confirm that no specific recommendation was made to cite any particular previously published work. Therefore, no additional citations have been added in response to such a request. Nevertheless, we have thoroughly examined our reference list to ensure its completeness, correctness, and relevance, and we have confirmed that no retracted articles are cited. If the editor or any reviewer recommends specific references for inclusion, we will gladly evaluate and incorporate them as appropriate. 4.“Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.” We sincerely appreciate the reviewer's meticulous feedback regarding the integrity of our reference list. In accordance with the comment, we have thoroughly re-examined all cited references and confirm that no retracted articles were included in the manuscript. Additional Editor Comments : 1. The consistency of LLMs in clinical decision-making processes was found to be low. This issue needs to be discussed in more detail. Response: We thank the reviewer for raising this critical issue. We have substantially expanded the discussion on the low consistency of large language models (LLMs) in clinical decision making, as requested. Specifically, we have added a detailed analysis in the Discussion section (please see lines 238–259 in the revised manuscript, with tracked changes). Lopez-Gonzalez et al.[24] reported only 34.16% concordance between GPT-4 and hospital algorithm decisions. Several factors contributed: ChatGPT-4’s knowledge is updated only until April 2023, it can only access open-access articles, and it explicitly states it is not designed for professional medical use. Sanchez-Cordero et al.[25] found that even after contextual training with 412 scientific articles, concordance improved only from 20.0% to 25.8%. Notably, ChatGPT tended to mirror the global “average” procedure distribution rather than tailoring recommendations to individual patients. The authors also noted that using a single center’s practice as the gold standard may introduce bias, as surgical choices vary across centers. Kahlon et al.[20] observed only fair agreement between ChatGPT-4 and bariatric surgeons, with moderate inconsistency across two runs. Surgeons can integrate nuanced, patient-specific information that AI cannot fully weigh, highlighting a key limitation. Jazi et al.[30] reported that ChatGPT-4 matched expert consensus in only 30% of complex cases, and gave inconsistent answers in 40% of scenarios. The model failed to recognize critical patient-specific risk factors. This issue extends beyond bariatric surgery. In summary, current LLMs lack individualized, context-aware reasoning; exhibit output instability; have outdated knowledge bases and limited access to full evidence; and cannot handle multi-criteria trade-offs. 2. The section on recommendations for future research could be more comprehensive Specifically, areas requiring further study should be identified. Response: We thank the reviewer for this valuable comment. We have revised and expanded the future research recommendations in the revised manuscript. In this expanded section, we clearly identify three interconnected priorities for future research .(please see lines 332–351 in the revised manuscript, with tracked changes). Specifically, future research should focus on three interconnected priorities. First, performance validation and multidimensional evaluation. Given the low consistency of LLMs in complex clinical decisions and their tendency to generate hallucinations or outdated information, rigorous and continuous evaluation and improvement are needed, including testing different models and prompt strategies. Special attention should be paid to reducing hallucinations, improving reproducibility, and enabling individualized recommendations. Assessment frameworks must systematically incorporate core metrics such as accuracy, reproducibility, readability, bias detection, and clinical utility, with standardized evaluation protocols. Second, development of human AI collaboration models. Research should use various data sources to explore optimal integration patterns that balance efficiency and safety. Different collaboration models should be tested continuously across diverse simulated clinical scenarios, evaluating their impact on decision quality, workflow efficiency, clinician acceptance, and patient outcomes. Third, establishment of ethical and legal governance frameworks. Clear operational guidelines are urgently needed: for data privacy, secure handling of patient queries and preventing unauthorized use of data for other purposes; for liability allocation, defining responsible parties when LLM suggestions contribute to adverse events and establishing a clear accountability framework; and for informed consent, disclosing AI use to patients. 3. Providing more detail on the ethical and safety risks associated with the use of LLMs would be beneficial in addressing concerns in this area. Response: We appreciate this valuable suggestion. We have added a dedicated, detailed paragraph on ethical and safety risks in the Discussion section. (please see lines 305–327 in the revised manuscript, with tracked changes). These technical limitations are closely intertwined with broader ethical and safety considerations, which deserve detailed elaboration. First, data privacy and confidentiality are critical concerns, as noted in multiple studies; patient queries often contain sensitive health information, and inputting such data into public or unencrypted LLM interfaces risks unauthorized access or data breaches, especially because some models store conversations for retraining, and the lack of transparency regarding how platforms handle personal data further exacerbates these risks. Second, algorithmic bias may lead to unfair or inaccurate recommendations; LLMs trained on non representative datasets may generate inappropriate recommendations, thereby exacerbating the risk of healthcare disparities. Third, liability and accountability remain unclear; the “black box” nature of LLMs makes their reasoning process opaque, and when an LLM contributes to a clinical decision that results in an adverse event, responsibility is ambiguous—it is unclear whether the developer, the hospital, the supervising physician, or the user should be held accountable, as current legal frameworks lack clear guidance for LLM assisted surgical decisions. Fourth, over reliance on LLMs may negatively affect the patient clinician relationship, as patients might reduce trust in human providers or delay necessary consultations. Finally, the most frequently cited issue is the “hallucination” phenomenon—LLMs tend to generate responses with high confidence and coherent structure even when the content is factually incorrect, which may mislead patients and lead uninformed users to adopt erroneous and potentially dangerous advice. Taken together, these risks emphasize that LLMs are not yet mature and must currently be positioned as supportive tools under human supervision, not autonomous advisors. Supplementary explanation: 1. NAAS will assess whether your figures meet our technical requirements by comparing each figure against our figure specifications. We thank the editor for the clarification regarding figure technical validation. We confirm full compliance with the figure formatting requirements through the following actions: ① Fig 1 (Fig1.tif) has been successfully validated via the NAAS pre flight process. ② The NAAS validated Fig1.tif will replace the original figure file in the resubmitted manuscript. 2. Data Availability Our study is a scoping review, not an original study generating new primary data. The minimal dataset required to replicate our findings—specifically, the key characteristics and results of the 21 included studies—is fully presented in Table 2 and the main text of the manuscript. No additional raw data were generated or analyzed beyond what is already reported. Accordingly, we have updated the Data Availability Statement in the manuscript to read: " All relevant data are within the paper and its Supporting Information files. As this is a review article, no new primary data were generated." We are grateful for the opportunity to improve our study through this revision. All changes have been thoroughly cross-checked for compliance with PLOS ONE’s guidelines, including figure formatting (validated via NASS), and reference integrity. Should additional clarifications or refinements be required, we remain fully committed to addressing them promptly. Thank you once again for your invaluable input. Sincerely, Jie Zheng Nursing College, Shanxi Medical University zhengjie@sxmu.edu.cn Tel: +86 18247758920 Attachment Submitted filename: Response to Reviewers.docx pone.0350748.s003.docx (34.8KB, docx) PLoS One. doi: 10.1371/journal.pone.0350748.r003 Decision Letter 1 Hongyang Ma Author information Copyright and License information Roles Hongyang Ma : Guest Editor © 2026 Hongyang Ma This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. PMC Copyright notice 19 May 2026 The Application of Large Language Models in Bariatric Surgery: A Scoping Review PONE-D-26-11116R1 Dear Dr. Jie Zheng, We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements. Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication. An invoice will be generated when your article is formally accepted. Please note, if your institution has a publishing partnership with PLOS and your article meets the relevant criteria, all or part of your publication costs will be covered. Please make sure your user information is up-to-date by logging into Editorial Manager at Editorial Manager® and clicking the ‘Update My Information' link at the top of the page. For questions related to billing, please contact billing support . If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org. Kind regards, Hongyang Ma, Ph.D., D.D.S Guest Editor PLOS One Additional Editor Comments (optional): Reviewers' comments: Reviewer's Responses to Questions --> Comments to the Author 1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.--> Reviewer #1: All comments have been addressed ********** -->2. Is the manuscript technically sound, and do the data support the conclusions? The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented. --> Reviewer #1: Yes ********** -->3. Has the statistical analysis been performed appropriately and rigorously? --> Reviewer #1: No ********** -->4. Have the authors made all data underlying the findings in their manuscript fully available? The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.--> Reviewer #1: Yes ********** -->5. Is the manuscript presented in an intelligible fashion and written in standard English? PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.--> Reviewer #1: Yes ********** -->6. Review Comments to the Author Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)--> Reviewer #1: (No Response) ********** -->7. PLOS authors have the option to publish the peer review history of their article ( what does this mean? ). If published, this will include your full peer review and any attached files. If you choose “no”, your identity will remain anonymous but your review may still be made public. Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy .--> Reviewer #1: Yes: Mustafa Cem Algin ********** -->--> -->--> PLoS One. doi: 10.1371/journal.pone.0350748.r004 Acceptance letter Hongyang Ma Author information Copyright and License information Roles Hongyang Ma : Guest Editor © 2026 Hongyang Ma This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. PMC Copyright notice PONE-D-26-11116R1 PLOS One Dear Dr. Zheng, I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS One. Congratulations! Your manuscript is now being handed over to our production team. At this stage, our production department will prepare your paper for publication. This includes ensuring the following: * All references, tables, and figures are properly cited * All relevant supporting information is included in the manuscript submission, * There are no issues that prevent the paper from being properly typeset You will receive further instructions from the production team, including instructions on how to review your proof when it is ready. Please keep in mind that we are working through a large volume of accepted articles, so please give us a few days to review your paper and let you know the next and final steps. Lastly, if your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org. You will receive an invoice from PLOS for your publication fee after your manuscript has reached the completed accept phase. If you receive an email requesting payment before acceptance or for any other service, this may be a phishing scheme. Learn how to identify phishing emails and protect your accounts at https://explore.plos.org/phishing . If we can help with anything else, please email us at customercare@plos.org. Thank you for submitting your work to PLOS ONE and supporting open access. Kind regards, PLOS ONE Editorial Office Staff on behalf of Dr. Hongyang Ma Guest Editor PLOS One Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials S1 Checklist. Preferred reporting items for systematic reviews and meta-analyses extension for scoping reviews (PRISMA-ScR) checklist. (DOCX) pone.0350748.s001.docx (769.9KB, docx) Attachment Submitted filename: Response to Reviewers.docx pone.0350748.s003.docx (34.8KB, docx) Data Availability Statement All relevant data are within the manuscript and its Supporting Information files. Articles from PLOS One are provided here courtesy of PLOS ACTIONS View on publisher site PDF (556.1 KB) Cite Collections Permalink RESOURCES Similar articles Cited by other articles Links to NCBI Databases Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top

Related documents

• Charakteristika und Behandlungsergebnisse von (teil-)stationär behandelten PatientInnen mit Bulimia nervosa / Anna Luisa Dieffenbacher ; Betreuer: Adrian Meule#1106548
• Variation in Large Language Model Recommendations in Challenging Inpatient Management Scenarios.#25849
• Multidisciplinary expert evaluation of large language models on questions regarding bariatric surgery: a comparative analysis of ERNIE Bot 4.0, ChatGPT-4, Claude 3 Opus, and Gemini Pro.#227959
• HerHealthEval: Evaluating Multilingual and Register-Sensitive Understanding of Women's Health Communication#1034484
• The effectiveness of whatsapp-based intervention among obese patients with type ii diabetes attending outpatient clinics, Hospital USM#613783