A publishing-ethics group also rules out an AI byline

The writer's problem

An editor at a journal that follows the Committee on Publication Ethics has to decide, in practice, whether a contributor's use of a chatbot changes who counts as an author of a paper and what that contributor must disclose. COPE's own position statement, Authorship and AI tools, last reviewed 13 February 2023, is the document COPE points members to on this question.

What the documents show

COPE states that AI tools cannot meet the requirements for authorship because they cannot take responsibility for the submitted work, and, as non-legal entities, cannot assert a conflict-of-interest position or manage copyright and license agreements. The statement requires authors who use AI tools in writing a manuscript, producing images or graphical elements, or collecting and analyzing data to disclose how and which tool was used in the Materials and Methods section, and it holds authors fully responsible for the content of their manuscript, including any parts an AI tool produced, making them liable for any breach of publication ethics that results. COPE frames this as joining other bodies that had already reached the same authorship conclusion, though this article does not independently verify each of those separate statements.

The editorial choice

For an editor building a submission policy on this position, the practical step is to require a named-tool disclosure in the specified section rather than a general mention elsewhere, and to treat an undisclosed AI contribution the same way any other undisclosed ghostwriting or unauthorized data source would be treated: as a potential ethics breach, not a technical oversight. That framing is an editorial extension of COPE's position, since the statement itself does not spell out a specific investigation procedure.

What stays with the author

Liability for the manuscript's content, including any AI-generated portion, rests with the named human authors under this position; a tool's fluency does not transfer any of that responsibility. Nothing on COPE's own site resolves how an editor independently confirms a disclosure is complete once it has been made, leaving that verification to each journal's own review process.

• Does the submission's Methods section name which tool was used and for what specific task?
• Would an undisclosed AI contribution here be treated the same as an undisclosed co-writer?
• Who at the journal checks a disclosure once filed, rather than simply filing it?

COPE's position adds a second organization's name to the same authorship line ICMJE draws, without adding a new test of its own.