Human Insight First: JAMA's New Guardrails on AI in Medical Publishing

Artificial intelligence (AI) is reshaping medical research and publishing at extraordinary speed. While it offers genuine efficiency gains, it has also produced fabricated references, formulaic submissions lacking clinical insight, and concerns about bias in generated images. The August 2026 JAMA editorial by Flanagin, Perlis, and Bibbins-Domingo responds to these challenges by updating author guidance. Low rates of AI disclosure, driven by fear of editorial bias or simple misunderstanding of policies, made clearer rules essential. Transparency, the authors emphasize, remains central to the integrity of the scholarly record.

What Remains Acceptable (with proper disclosure)?

The updated guidance permits several practical uses of AI. Research applications are allowed when fully described in the Methods section. Literature searching, assistance with manuscript preparation, translation, and generation of data visualizations are similarly acceptable, provided they are disclosed in the Acknowledgment section or figure legend and authors verify accuracy. Routine grammar and spelling checks require no disclosure. These permissions recognize AI’s legitimate value while insisting that human authors retain full responsibility for every claim, reference, and conclusion.

What Is Not Permitted?

Certain applications are now explicitly prohibited. AI may not be used to draft Opinion articles, Letters to the Editor, or online Comments; human insight is prioritized. Generation or formatting of bibliographic references is strongly discouraged because even advanced models produce realistic but non-existent citations. Clinical images, illustrations, video, and audio created or manipulated by AI are barred unless they form part of a formal research design that is fully disclosed. Entering manuscripts into external tools for peer review or drafting responses to reviewers violates confidentiality and is forbidden. These restrictions address documented problems of low-value, formulaic content and threats to the reliability of the published record.

Disclosure Requirements and Author Accountability

When AI is used for content creation, revision, or formatting, authors must report the software name, version, manufacturer, dates of use, a brief description of how and where it was applied, and an explicit statement that they take responsibility for accuracy and integrity. Failure to disclose can result in rejection or post-publication correction. The editorial reaffirms a foundational principle: authors remain fully accountable for all submitted content, whether AI-assisted or entirely human-generated. Disclosure alone cannot eliminate fabricated material; responsibility and editorial vigilance must operate together.

Indian journals and academic institutions would benefit from adopting similarly clear, transparent policies. Medical colleges, research societies, and national bodies such as the National Medical Commission (NMC) could incorporate structured training on responsible AI use into research methodology curricula. Authors should treat disclosure not as a bureaucratic hurdle but as a professional responsibility that protects their credibility and the trustworthiness of Indian research on the global stage. Career advancement systems that heavily weight publication quantity must also evolve to reward quality, originality, and ethical practice over volume alone.

Implications for Authors and the Future of Medical Publishing

For clinicians, researchers, and educators who contribute to the literature, the practical message is straightforward. Use AI thoughtfully for efficiency in drafting, searching, or visualization, but never as a substitute for expertise or original insight. Verify every reference and image. Reserve Opinion and correspondence pieces for genuine human perspectives. The guidance will continue to evolve as models improve and new risks appear. In the meantime, JAMA Network’s clarity offers a useful model for other journals. By combining pragmatic permissions with firm boundaries, the editorial helps preserve the trustworthiness of medical knowledge at a time when that trust is more critical than ever.

Reference

Flanagin A, Perlis RH, Bibbins-Domingo K. Updated Guidance for Author Use of AI in Medical Publication. JAMA. Published online August 10, 2026. doi:10.1001/jama.2026.16613