From Virtual Tutors to Professional Identity: Generative AI and Large Language Models in Medical Education

Abstract

Background: Generative artificial intelligence (AI), particularly large language models (LLMs), is rapidly transforming medical education. These tools can act as virtual tutors, generate teaching materials and support clinical simulations, but they also raise concerns regarding accuracy, bias, academic integrity, data protection and professional identity formation. Objective: This narrative review aims to synthesise current knowledge on the use of generative AI - with a focus on LLMs - in medical education, identify key benefits and risks, and discuss implications for curricula and professionalism in medicine and other health professions. Methods: Publications from 2021–2025 were included if they addressed applications of generative AI/LLMs in medical or health professions education, or discussed their impact on curricula, assessment, professionalism or future physician competencies. Original empirical studies, reviews, conceptual papers and policy/guideline documents were included. Data were extracted and synthesised narratively according to four domains: areas of application, benefits, risks/limitations and recommendations/frameworks. Results: Generative AI is being used to support self-directed learning (on-demand explanations, personalised practice questions), clinical simulations and virtual patients, faculty work in content generation and assessment, and research in medical education. Reported benefits include increased accessibility, personalisation, scalability and opportunities to practise clinical reasoning and communication in low-stakes environments. Major risks comprise hallucinations and bias, threats to academic integrity, potential deskilling and over-reliance, and privacy and confidentiality concerns.