Prevalence & utilisation of AI chatbots for medical education among students of medical colleges

Sir,

Medical education requires high cognitive engagement and rapid assimilation of knowledge. Artificial intelligence (AI) chatbots act as personalized learning assistants, especially during self-directed learning, helping students generate clinical scenarios, receive simplified explanations, and summarise difficult topics. A national study from China showed that although only 28.72 per cent of students actively used ChatGPT, most recognised its potential to improve efficiency and understanding in students’ learning¹. Similarly, a study from the USA highlighted ChatGPT’s popularity for aiding clinical reasoning and exam preparation². In Saudi Arabia, awareness was high, with greater use among female and fourth -year students, though concerns remained about accuracy³.

A mixed-methods study in the UK revealed cautious optimism but concerns over ethical user and reliability of AI Chatbots persisted⁴. In India, medical students valued AI for summarising vast content and exam preparation but noted a need for critical appraisal⁵. Innovative models such as ChatGPT-assisted problem-based learning (PBL) improved engagement and comprehension⁶. Educators agreed that AI cannot replace mentorship but can augment efficiency and retention⁷. Systematic reviews have emphasized ChatGPT’s role in medical education and digital health, nothing benefits like flexibility and instant feedback, but also factual inconsistencies^8,9. Given this global shift, it becomes essential to assess how medical students in different regions perceive and utilize AI chatbots. Understanding their prevalence, usage, and perceived benefits or limitations can inform strategies for integrating AI into undergraduate curricula.

The research was conducted under the department of Forensic Medicine and Toxicology, Hinduhridaysamrat Balasaheb Thackeray Medical College and Dr. R. N. Cooper Municipal General Hospital, Mumbai, Maharashtra, India, involving medical students from all academic years. The institution’s strong digital infrastructure and widespread internet access among students provided a suitable environment for investigating AI chatbot utilisation.

A cross-sectional design was employed to assess awareness, usage, and attitudes toward AI chatbots data were collected through a structured questionnaire covering demographics, familiarity, frequency and purpose of use, perceived benefits, limitations, and willingness to integrate chatbots into learning. The study was conducted from June 10, 2025 to July 21, 2025. The study population comprised of 385 undergraduate MBBS students aged 18-24 yr studying at the institution. Students from all years were included to ensure diversity. Sample size was calculated using standard statistical formulas, assuming 30 per cent prevalence, 95 per cent confidence interval, and five per cent margin of error, with 10–15 per cent oversampling to account for incomplete responses.

Undergraduate MBBS students aware of AI chatbots (e.g., ChatGPT) who gave informed consent were included in the study. On the other hand, students were excluded if they had never heard of AI chatbots, declined consent, or provided incomplete responses.

A self-administered structured questionnaire, developed after literature review and expert consultation, was distributed electronically via Google Forms, accessible on smartphones and computers. Reminders through institutional mailing lists and social media enhanced participation while ensuring anonymity and confidentiality. Data were analysed using SPSS and R software. Descriptive statistics summarised demographics and usage. Chi-square tests assessed associations, while logistic regression identified predictors of chatbot use, adjusting for confounders.

A total of 385 MBBS students from various academic years were included in this study on awareness, accessibility, and utilisation of AI chatbots. Students from all MBBS years were included, with higher participation from third- and final-year students. Males (54.8%) slightly outnumbered females (45.2%).

Most students were comfortable with digital tools, and over 90 per cent reported reliable access to the internet and devices. However, only 35.6 per cent were aware of prompt engineering, with most using brief prompts, reflecting limited engagement depth. Key barriers included lack of technical skills, inadequate awareness of tools, and occasional internet/device constraints. Chatbots were considered most beneficial for quick information access, personalised training, practice with case scenarios, and research support, highlighting their versatile role in learning.

Overall, 97.7 per cent were aware of AI chatbots, and 86 per cent had used them for academics purpose. Usage frequency varied from daily to occasional, showing widespread but uneven engagement. The most common applications were exam preparation (69.4%), research support (47.5%), and mnemonic creation (28.8%). Multiple choice questions (MCQ) generation was less common. AI chatbots were chiefly used for exams, research, and mnemonics, with limited use for MCQ generation. Smartphones were the most frequent access device, followed by tablets and desktops. Students also used other digital tools such as lecture videos, e-books, and MCQ banks.

Over 60 per cent viewed chatbots as valuable and supported their curricular inclusion, while opinions on regulation were divided. Most students rated their experience positively, though views on productivity benefits were mixed. Many endorsed chatbot use as a supplement to traditional methods and a facilitator of collaborative learning.

In this cross-sectional study of 385 MBBS students, awareness (97.7%) and adoption (86%) of AI chatbots were high. While most students were comfortable with technology, only 35.6 per cent knew prompt engineering. Chatbots were mainly used for exam preparation and knowledge retrieval, with barriers including limited skills and concerns about reliability. Perceptions were positive, with support for curricular integration, though regulation and productivity views were mixed.

Findings are consistent with international studies. Tangadulrat et al¹⁰ reported 83.1 per cent positive perceptions among medical students, and Moskovich and Rozani⁴ found widespread acceptance but accuracy concerns. Studies in Europe and the U.S. reported similar adoption for exam preparation¹¹, while advanced uses remain limited. ChatGPT has shown performance comparable to third-year medical students on United States Medical Licensing Examination questions¹².

Despite such promise, knowledge gaps remain. The results underscore the need for training in prompt engineering and responsible AI use¹³. Strengths include a large, diverse sample and structured tool. Limitations are self-reported data, single-institution scope, and cross-sectional design. Medical curricula should include AI literacy, covering prompt design, appraisal of outputs, and ethical considerations¹⁴. Clear policy guidelines are needed given mixed views on regulation. Longitudinal and interventional studies, alongside qualitative exploration, are required to evaluate outcomes and guide responsible AI integration in medical education.