Answers are grounded in this paper and cite their sources.
INTRODUCTION: ChatGPT has attracted growing interest in medical education, but the knowledge structure and thematic development of the associated literature remain insufficiently characterized. This study aimed to map the evolution of ChatGPT-related research in medical education and interpret its thematic patterns through a socio-technical framework. METHODS: A bibliometric and knowledge-mapping analysis was conducted using 637 English-language articles and reviews on ChatGPT indexed in the Web of Science Core Collection and published between 2023 and 2025. CiteSpace was employed to analyze publication trends, collaboration networks among authors, institutions, and countries, keyword co-occurrence, thematic clusters, and keyword burst patterns. Algorithmically generated clusters were interpreted using three author-developed analytical dimensions. The dataset was not considered representative of the broader generative AI literature. RESULTS: Annual publication output increased from 123 in 2023 to 240 in 2024 and 274 in 2025. Ten keyword clusters were mapped to technological efficacy (three clusters), pedagogical transformation (six), and ecosystem governance (one), indicating uneven empirical representation. Keyword patterns suggested a broadening of attention from performance benchmarking to pedagogical applications, human-AI interaction, and governance concerns. Burst findings were interpreted as exploratory changes within the three-year window. Collaboration network analysis revealed uneven research participation and dispersed collaboration within the retrieved literature. CONCLUSION: The findings suggest that research on ChatGPT in medical education is expanding beyond isolated performance evaluations to consider its relationships with teaching, learning, and governance. The framework provides an interpretive synthesis of these research patterns, rather than a validated developmental model or evidence of educational effectiveness. Its applicability to other generative AI systems requires further investigation.
No takes yet. Share an insight, caveat, or question.
Hu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: