Artificial intelligence use, perceptions, and educational needs among medical undergraduates at a Sri Lankan medical faculty: a cross-sectional survey

Abstract

Artificial intelligence (AI) is rapidly transforming medicine, yet AI competencies remain insufficiently integrated into medical curricula, particularly in low- and middle-income countries. This study investigated AI tool use, self-rated proficiency, and attitudes towards AI in medical education, physician roles, and patient care among medical undergraduates at a Sri Lankan medical faculty, including comparisons across year groups and gender. A cross-sectional online survey was conducted among Years 1–5 medical undergraduates at SUSL in May 2025. The questionnaire assessed demographics, AI tool use, self-rated proficiency, and 12 Likert-scale items covering curriculum integration, physician roles, and patient care. Descriptive statistics, Kruskal–Wallis H-tests, Mann–Whitney U-tests, and chi-square tests were used for analysis. A total of 369 responses were analysed. Reported use of AI-based applications was high (99.2%; 366/369). Among respondents reporting AI use ( n = 366), ChatGPT was the most frequently used tool (98.1%), followed by Google Gemini (35.0%), DeepSeek (20.2%), and Microsoft Copilot (6.3%). The main uses of AI were subject-matter reference (89.3%), general information retrieval (73.0%), assignment assistance (50.8%), and writing improvement (39.6%). Self-rated proficiency was highest for ChatGPT, with 92.4% of respondents rating their proficiency as good or higher. Among respondents in the medical information/reference application analysis ( n = 360), 77.5% used Medscape, followed by UpToDate (37.2%) and WebMD/WebMD Symptom Checker (15.3%). Most students agreed that AI could enhance learning in preclinical (85.1%), paraclinical (77.8%), and clinical (68.6%) subjects. After adjustment for multiple comparisons, support for including AI in the medical curriculum differed significantly across year groups (H = 29.00, Holm-adjusted p < 0.001), with higher agreement among some earlier-year groups. Medical undergraduates demonstrated high levels of self-reported AI tool use despite limited formal training. Attitudes toward AI differed for selected items between preclinical and clinical students, although longitudinal change cannot be inferred from these cross-sectional data. These findings support consideration of structured, ethically grounded AI education in undergraduate medical curricula to address students’ educational needs relating to the safe, effective, and responsible use of AI.