Medical Students’ Attitudes, Perceptions, and Usage of Large Language Models in Education: A Questionnaire-based Study

Abstract

Objectives: This study examined medical students’ attitudes, perceptions, and usage patterns of AI, particularly large language models (LLMs), in medical education. The goal was to explore how these tools are used for academic purposes and their potential integration into medical curricula.

Methods: A cross-sectional questionnaire was distributed to medical students across six academic years at a Portuguese institution during autumn 2024. Respondents rated their study habits, the relevance of digital resources, frequency of engagement with LLMs, trust in AI-generated content, and opinions on curricular integration.

Results: A total of 306 students (13.4% response rate) completed the survey. AI was used by 87% of respondents, primarily for resolving theoretical doubts (84%), while its application in complex academic tasks was limited. Freely available models (GPT-3.5) were the most commonly used, whereas only 17% had experience with paid versions such as GPT-4. Trust in AI-generated clinical recommendations was low, with only 16% considering them reliable in a clinical case-based scenario. Limited familiarity (69%) and cost (58%) were identified as key barriers to broader adoption. No substantial evidence suggested widespread use of AI for academic misconduct. Despite scepticism regarding its reliability in clinical contexts, most respondents supported AI integration into the curriculum, with 65% favouring an optional course.

Significance: Students frequently use AI for theoretical learning but remain sceptical of its reliability in medical decision-making. Addressing concerns through AI literacy and reducing cost barriers may encourage responsible adoption in medical education.