Authorship · Reference
Can an AI be an author?
Generative-AI tools cannot be listed as authors of a research output: the major editorial bodies agree that an author must be able to take responsibility and be accountable for the work, which an AI cannot do — but authors must disclose substantive AI use.
The step most authors miss
Doing CRediT right? Don’t stop at the statement.
A CRediT statement credits you inside one paper. The recognition CRediT was built for happens when those roles are tied to you, persistently. Sign in with your ORCID — free — and claim your CRediT contributions on casrai.org, the home of the standard. They become a verified, portable part of your identity, not a line that disappears into one PDF.
Free: claim your contributions, then export a journal-ready CRediT statement, schema.org structured data, JATS XML, CSV or BibTeX — and preview your public profile. A membership publishes that profile publicly and verifies the journals you serve.
Why an AI cannot be an author
The leading editorial bodies — the ICMJE, COPE and major publishers including Nature’s publisher — converged quickly on the same position: a large language model or other AI tool cannot be an author. Authorship entails accountability for the work, agreement to be answerable for its accuracy and integrity, and the ability to approve the final version and manage conflicts of interest. An AI system cannot meet, consent to or be held to any of these responsibilities.
This is not a statement about how capable AI tools are; it is about the nature of authorship. Even where an AI tool has materially helped produce text, analysis or figures, responsibility for the output still rests entirely with the human authors who chose to use it.
Disclosure, not authorship
The agreed approach is disclosure. Authors who use generative AI substantively — for drafting text, generating or editing images, analysing data, or producing code — should describe that use transparently, typically in the methods or acknowledgements, naming the tool and explaining how it was used. The ICMJE and COPE both ask authors to disclose AI use and to take full responsibility for the accuracy and integrity of any AI-assisted content, including checking for fabricated references or errors. Routine assistance such as basic spelling and grammar checking generally does not require disclosure.
What authors remain responsible for
Because the human authors are accountable, they must verify everything an AI tool contributes: confirm that cited references exist and are accurate, check that generated data or analyses are correct, and ensure no plagiarised or fabricated material has entered the manuscript. Authors also retain responsibility for any conflicts of interest and for compliance with the journal’s AI policy, which increasingly forms part of the submission declarations.
A fast-moving policy area
AI disclosure policy is evolving quickly, and individual journals and publishers set their own specific requirements on top of the ICMJE and COPE baselines. Before submitting, authors should check the target journal’s current AI policy, since the level of detail required and the section in which disclosure must appear vary. The underlying principle, however, is stable: AI tools are disclosed as tools, never credited as authors.
Key facts
At a glance
- Consensus: AI tools cannot be listed as authors
- Reason: authorship requires accountability only a person can hold
- Bodies: ICMJE, COPE and major publishers agree
- Requirement: disclose substantive AI use (tool and how it was used)
- Author duty: verify AI output — references, data, no fabrication or plagiarism
- Exempt: routine spelling/grammar assistance usually needs no disclosure
Common questions
FAQ
Can an AI be listed as an author?+
No. The ICMJE, COPE and major publishers agree that AI tools cannot be authors, because authorship requires accountability and final approval that only a person can provide.
Do I have to disclose using ChatGPT or other AI tools?+
Yes, where the use was substantive — for drafting text, generating images, analysing data or producing code. Disclose the tool and how it was used, usually in the methods or acknowledgements. Routine spelling and grammar checking generally does not need disclosure.
Who is responsible for AI-generated content in a paper?+
The human authors are fully responsible. They must verify that references exist, that data and analyses are correct, and that no fabricated or plagiarised material has been introduced by the AI tool.
Why can’t an AI take credit for the work it does?+
Authorship is not only about producing content; it requires agreeing to be accountable for the work and approving the final version. An AI cannot consent to or be held to those responsibilities.
Where should AI use be disclosed?+
Typically in the methods or acknowledgements section, but journals set their own requirements, so check the target journal’s AI policy before submitting.
