Usage pattern of artificial intelligence and evidence-based platforms among medical undergraduates at Sabaragamuwa University of Sri Lanka

Abstract

In the field of Medicine, evidence-based platforms such as Medscape and UpToDate provide reliable, validated clinical information. Despite that, artificial Intelligence (AI) is increasingly being used in medical education and clinical practice, supporting both learning and decisionmaking. General AI tools like ChatGPT, Google Gemini, Microsoft Copilot, and DeepSeek are trending among medical students due to their easy accessibility and convenience, despite content reliability. This study observed the prevalence of AI tools and medical tools usage among undergraduate medical students at the Faculty of Medicine, Sabaragamuwa University of Sri Lanka. A cross-sectional descriptive survey was conducted in May and June 2025 using a structured online questionnaire distributed to all seven MBBS (Bachelor of Medicine, Bachelor of Surgery) batches. A total of 372 valid responses were analysed in Microsoft Excel, SPSS version 28, and McNemar’s tests were applied to compare the usage of general-purpose AI tools with medical-specific platforms within each batch. To account for multiple comparison Bonferroni correction was applied (adjusted significance p < 0.0071). ChatGPT was the most widely used general AI tool (94.4%-100%), followed by Google Gemini (20%-56%), Microsoft Copilot, and DeepSeek. Among medical-specific tools, Medscape was most frequently used (55.6%- 96.7%), followed by UpToDate (6.6% - 86.7%). Overall, McNemar’s test showed a significant difference between general AI and medical-specific tool usage (χ² = 29.43, p <0.001), with general AI used more often. Batch-level analysis found no significant differences in the most senior clinical batches (Batch 01: χ² = N/A - zero discordant pairs; Batch 02: χ² = 1.00, p = 0.317; Batch 03: χ² = 1.00, p = 0.317; Batch 04: χ² = 0.00, p = 1.00; Batch 05: χ² = 3.00, p = 0.0832). Significant differences were observed in junior pre-clinical batches (Batch 06: χ² = 8.33, p = 0.00389; Batch 07: χ² = 18.00, p< 0.001), indicating higher use of general AI tools over medical-specific platforms in earlier training years. These results show extensive AI tools usage among Sabaragamuwa medical undergraduates, especially in the preclinical period. Students in clinical stages reported more balanced use of both general and medical-specific tools. The findings highlight the need for a structured AI module within the MBBS curriculum, which will help to guide the students in AI use for medical education and clinical decision-making