Charting the rise of artificial intelligence in medical education: a global bibliometric and network analysis, 2000–2026

Abstract

Artificial intelligence has moved from a peripheral curiosity to a core feature of modern medical education through adaptive learning, simulation-based training, automated assessment, and large language models such as ChatGPT. Publication activity has accelerated rapidly, but the structure of this expanding literature has not been comprehensively mapped. Bibliographic data were obtained from the Web of Science Core Collection using a search strategy that combined medical education terms in the title field with artificial intelligence terms in the topic field. Articles, review articles, and data papers published from 2000 to 2026 were included. A total of 2,298 records were retrieved. Microsoft Excel was used for descriptive analyses, and VOSviewer (version 1.6.20) was used to construct keyword co-occurrence, country collaboration, and institutional collaboration networks. Data for 2026 represent a partial calendar year. Publication output remained limited until the late 2010s and rose sharply after 2021. Of the 2,298 records, 2,066 (89.9%) were published from 2021 to 2026, with the highest complete-year output in 2025 (821 publications). The United States led in publication volume (918 records) and raw citation count (14,582 citations), followed by the People’s Republic of China (367 records; 4,404 citations). The Journal of Surgical Education published the most articles. Harvard Medical School led institutional output, whereas Stanford University and the University of California San Francisco had higher raw citation totals despite fewer publications. Artificial intelligence, medical education, ChatGPT, medical students, and machine learning were the most prominent keywords. Artificial intelligence in medical education is a rapidly expanding but geographically and institutionally concentrated field. Raw citation patterns differed from publication volume, although these unadjusted measures should not be interpreted as direct comparisons of research quality. Thematic patterns emphasize learner-facing applications, assessment, and generative AI, while responsible curricular integration and empirical evaluation remain important priorities. These findings are especially relevant to emergency medicine education, where simulation, time-sensitive decision-making, and competency assessment are central.