The Presence and Nature of AI-Use Disclosure Statements in Medical Education Journals: A Bibliometric Study

BackgroundAs AI-use becomes more common in research, disclosure policies have emerged to ensure transparency and appropriateness. However, database research in other fields suggests that disclosure may lag behind AI-use. Medical education journal editors report that submitted manuscripts rarely include AI-use disclosures, and they perceive a lack of clarity regarding when and how AI-use should be disclosed. However, we lack objective evidence regarding the incidence and nature of AI-use disclosure in medical education. MethodsUsing bibliometric methods, we searched a database of 24 leading medical education journals for articles published between January and July 2025 (n=2,762 articles). Screening with Covidence software excluded 716 non-empirical and/or non-English language articles. The remainder (n=2,046) were examined for the presence of AI-use disclosures, which were content-analyzed. Results2.5% of empirical articles (n=51) had an AI disclosure statement. BMC Medical Education contained the most disclosures (24), followed by Medical Teacher (7) and Journal of Surgical Education (4). Forty-two articles were authored in non-native English-speaking countries, and 69.4% of all first authors had begun publishing in the past decade. Disclosures averaged 43 words and described use superficially: most commonly "editing" and "translation". Of 18 named tools, ChatGPT was most common. Most disclosures explicitly attested to author responsibility for AI-produced material. Disclosures usually appeared in acknowledgements; those located in methods lacked responsibility attestation. Negative disclosures attesting that AI was not used were also present. DiscussionAI-use disclosures in medical education journals are rare and appear mostly in work from non-native English-speaking regions of the world. A shared disclosure practice is evident: name the tool and affirm author responsibility, but describe use superficially. This suggests a practice of "safe" disclosure that may be more performative than informative, therefore failing to satisfy the goal of ensuring transparent and ethical AI use in research.