ICMJE Generative AI Policy

The ICMJE Generative AI Policy refers to the guidance the International Committee of Medical Journal Editors added to its Recommendations for the Conduct, Reporting, Editing, and Publication of Scholarly Work in Medical Journals in the May 2023 update, at Section II.A.4. Because ICMJE’s Recommendations are followed or referenced by thousands of biomedical and health-science journals well beyond the committee’s own member journals, this is one of the most widely cited reference points for how generative AI use is handled in manuscript submission — distinct from a single journal or publisher’s own house policy such as the Nature Portfolio AI Policy or the IEEE Generative AI Policy, both of which build on the same underlying norms but are written and enforced independently by those publishers.

No AI tool can be listed as an author

ICMJE states plainly that chatbots such as ChatGPT should not be listed as authors because they cannot be responsible for the accuracy, integrity, and originality of the work, and cannot give the approvals ICMJE’s own authorship criteria require. This is not a judgment about how useful or capable a tool is — it follows directly from ICMJE’s four-part authorship test, under which anyone credited as an author must be able to be held accountable and to approve the final version submitted for publication. A software tool cannot hold accountability in that sense, so it cannot qualify as an author no matter how substantial its contribution to the drafting. This is the same reasoning behind the related, narrower term AI co-authorship rejection (ICMJE 2023), which documents this specific rule as it was introduced.

Disclosure is required, and the location depends on how the tool was used

ICMJE requires that, at submission, journals ask authors to disclose whether they used AI-assisted technologies — including large language models, chatbots, or image generators — in preparing the work. Where that disclosure belongs depends on what the tool was used for: assistance with writing, editing, or proofreading text is described in the Acknowledgments section, while use of AI to collect or analyze data, or to help generate figures, is reported in the Methods section instead. This split matters for authors assembling a submission checklist, since a disclosure placed in the wrong section can read as incomplete or inconsistent to an editor even when the underlying use was properly reported. See CASRAI’s general Generative-AI disclosure statement entry for how publishers commonly implement this as a standalone, labelled section of the manuscript.

Authors remain accountable for AI-assisted content

ICMJE is explicit that humans are responsible for any submitted material that involved the use of AI-assisted technologies, and that authors should carefully review and edit AI-generated output because such tools can produce authoritative-sounding text that is incorrect, incomplete, or biased. In practice this means an author cannot treat an AI tool’s output as a source of fact-checked citations, data summaries, or claims — every AI-assisted sentence carries the same verification burden as one the author wrote unaided, and responsibility for any resulting error sits with the named human authors, not the tool.

How this relates to journal and publisher policies

ICMJE’s Recommendations are guidance a journal can choose to follow or adapt, not a binding regulation, so individual publishers still publish their own AI policies that reference or extend it — see CASRAI’s guide to the Journal and Publisher Policies on Generative AI in Manuscripts for how major publishers compare, and the Can AI Be Listed as an Author? guide for how ICMJE, COPE, and individual publishers line up on the authorship question specifically.

References

• ICMJE Recommendations, Section II.A.4, “Non-Author Contributors” (updated May 2023) — icmje.org
• COPE Position Statement on Authorship and AI Tools (February 2023)