Current status of medical students’ use of large language models for graduation thesis writing: a comparative cross-sectional survey

Abstract

Objective Guided by the Technology Acceptance Model (TAM) as a conceptual framework, this study explored medical students’ use of large language models (LLMs) in graduation thesis writing, focusing on application patterns, perceptions, and potential risks. Methods A comparative cross-sectional questionnaire survey using convenience sampling was conducted among medical graduates from medical schools in China in 2024 and 2025. The self-designed 25-item questionnaire included five-point Likert items on perceived advantages and disadvantages of LLMs. Internal consistency was assessed using Cronbach's α , exploratory factor analysis (EFA) was performed, and univariable and multivariable logistic regression analyses were used to identify factors associated with LLMs use. Results A total of 341 valid questionnaires were collected. Overall, 54.25% (185/341) of respondents reported using LLMs in graduation thesis writing, with use reported by 27.37% (52/190) in 2024 and 88.08% (133/151) in 2025. ChatGPT was the most frequently used platform in 2024 (82.69%, 43/52), whereas Deepseek was the most frequently used platform in 2025 (88.72%, 118/133). LLMs were used primarily during the literature review stage (74.05%, 137/185), mainly to provide writing suggestions (82.16%, 152/185). Cronbach's α was 0.720 for perceived advantages and 0.650 for perceived disadvantages, and exploratory factor analysis identified a two-factor clustering pattern consistent with the predefined grouping of perceived advantages and disadvantages. Among LLMs users, 68.65% (127/185) believed that LLMs use would not compromise academic integrity, whereas 47.57% (88/185) remained concerned about potential academic misconduct. The main perceived disadvantages included the risk of academic misconduct, limited content accuracy, and data privacy and security concerns. The findings were discussed with reference to TAM-related concepts as an interpretive framework. Conclusion LLMs use was substantially higher in the 2025 survey than in the 2024 survey. Although LLMs improve writing efficiency and provide valuable writing support, concerns regarding academic integrity, content accuracy, and data privacy remain important considerations. The TAM provides a useful framework for interpreting students’ adoption patterns and may inform the development of AI education and governance strategies in medical education.