Three publishing bodies agree AI cannot be an author

The writer's problem

A researcher who used a chatbot to help draft part of a manuscript, or an editor deciding how to handle such a submission, runs into the same question across medical and academic journals: can the tool be credited, and what has to be disclosed if it is not? Three influential publishing-ethics bodies, the International Committee of Medical Journal Editors (ICMJE), the World Association of Medical Editors (WAME), and the Committee on Publication Ethics (COPE), have each published their own answer.

What the documents show

All three converge on the same bar. ICMJE's guidance states that chatbots 'should not be listed as authors because they cannot be responsible for the accuracy, integrity, and originality of the work,' and requires authors to disclose at submission whether they used such tools and how, in the cover letter and manuscript. COPE's position statement, last reviewed 13 February 2023, states 'AI tools cannot meet the requirements for authorship as they cannot take responsibility for the submitted work,' adding that as non-legal entities they 'cannot assert the presence or absence of conflicts of interest nor manage copyright and license agreements,' and requires disclosure in the paper's methods section. WAME's recommendations, revised 31 May 2023, state as their first principle that 'only humans can be authors,' and that authors remain 'responsible for the material generated by the chatbot,' including checking it against original sources. The three differ mainly in where disclosure belongs, WAME in an acknowledgment with prompts, ICMJE in the cover letter and manuscript, COPE in methods, rather than in the underlying authorship rule.

The editorial choice

A journal or manuscript editor enforcing this rule is applying a shared industry norm, not one organization's idiosyncratic policy, which strengthens the case for treating non-disclosure as a real problem rather than a technicality. This is an editorial position the three bodies share explicitly: responsibility for a manuscript's accuracy cannot be delegated to a tool, whatever role that tool played in producing a draft.

What stays with the author

None of the three documents specifies a penalty scale for undisclosed AI use beyond general statements that it may be treated as misconduct; enforcement is left to individual journals. Nor do they resolve harder edge cases, such as AI-assisted data analysis presented without any generated text at all. The responsibility the three organizations describe, for accuracy, originality and the manuscript's claims, remains the human author's regardless of which tool touched the draft.

• Has the specific AI tool and its role been disclosed in the location this journal's own policy specifies?
• Would a reviewer, given the disclosure, be able to independently judge which claims came from the author's own analysis?
• Does the journal being submitted to follow one of these three bodies' guidance, or its own separate policy?

The agreement across ICMJE, COPE and WAME is notable precisely because these bodies otherwise serve different fields and functions; on this one point, academic publishing has settled faster than book publishing has.