ChatGPT-related research in medical education: A dual-database bibliometric review and science-mapping analysis, 2023 to 2026

Background: The medical-education literature on ChatGPT has grown rapidly since 2023, but bibliometric patterns vary with database coverage, indexing fields, and the boundary used to define educational relevance. We examined publication growth, collaboration, and thematic development and assessed whether the leading patterns were stable under a stricter topic definition. Methods: Web of Science Core Collection (WoSCC) and PubMed were searched on September 5, 2026; the PubMed publication-date filter extended through September 4, 2026. WoSCC supported citation- and Author Keyword-based science mapping, whereas PubMed provided a complementary Medical Subject Headings-based view. The expanded corpora contained 665 WoSCC and 1106 PubMed records, with strict topic-specific subsets of 532 and 898 records, respectively. Bibliometrix/R, VOSviewer, CiteSpace, and Microsoft Excel were used for descriptive bibliometrics, collaboration networks, thematic mapping, multiple correspondence analysis, keyword co-occurrence, Trend Topics, burst detection, and journal mapping. Data from 2026 were treated as partial-year observations. Results: WoSCC publication output increased from 66 records in 2023 to 169 in 2024 and 242 in 2025; 188 records were indexed during the partial 2026 period. The United States had the largest corresponding-author output, and Sichuan University was the leading displayed affiliation. The keyword structure centered on medical education, ChatGPT, artificial intelligence, and large language models. CiteSpace identified 9 keyword clusters, including medical education, large language models, medical students, generative artificial intelligence, prompt engineering, clinical decision-making, and patient education. Recently emerging terms included DeepSeek, clinical reasoning, and anatomy education. Author Keyword rankings were highly concordant between the expanded and strict WoSCC datasets (Spearman ρ = 0.973, P < .001). Conclusion: ChatGPT-related medical-education research expanded substantially after 2023 and diversified across assessment, clinical reasoning, educational applications, and broader LLM-related themes. The leading WoSCC keyword structure remained stable under the stricter topic definition. WoSCC and PubMed captured different bibliographic features and are best interpreted as complementary sources.