Two Decades of Artificial Intelligence in Medical Education: Progress, Promise, and Persistent Gaps – A Narrative Review

Artificial intelligence (AI) has transitioned over the past two decades from a set of experimental tools to an integral component of medical education. Despite several recent reviews, a consolidated two-decade overview that synthesises thematic developments and persistent gaps remains limited. This narrative review therefore examines how AI has shaped medical education between 2005 and 2025, highlighting progress, evolving practices, and areas requiring further research. A narrative review methodology was used to synthesise peer-reviewed literature published from January 2005 to May 2025. Searches were conducted in PubMed and Google Scholar. From the search and screening process, approximately 63 articles were included for thematic synthesis. These were categorised into five domains: teaching and instruction, assessment and evaluation, clinical simulation and training, curriculum development, and lifelong medical learning. Across the 63 included studies, AI adoption in medical education expanded significantly after 2018. Evidence demonstrated improvements in learner engagement, personalised instruction, automated scoring of clinical reasoning, AI-enabled simulation with adaptive feedback, and emerging curriculum initiatives introducing AI competencies. However, most findings were short-term, with limited longitudinal outcomes, heterogeneous methodological quality, and a lack of standardised curricular frameworks. AI has become a transformative force across multiple dimensions of medical education. While offering important opportunities for personalised learning, scalable assessment, and immersive simulation, its integration requires careful attention to ethical, pedagogical, and equity considerations. Further research is needed to establish robust long-term outcomes, develop standardised curricula, and promote equitable access to AI-enhanced education.