In short: Generative AI can assist with research writing, but it cannot be an author, its substantive use must be disclosed, and you — the human author — remain fully responsible for accuracy, originality and integrity. The biggest practical risk is fabricated citations.
What AI can and cannot do
Generative AI tools can help with brainstorming, structuring, language editing and summarising. They cannot take responsibility for the work, guarantee accuracy, or replace your judgement. That distinction drives every rule below and is why our editorial policies and research integrity standards treat AI as a tool, not a contributor.
The rules
1. AI cannot be an author
Authorship requires accountability and the ability to approve the final work — things an AI tool cannot do. Guidance from the ICMJE and COPE is clear that AI tools must not be listed as authors.
2. You remain responsible
Human authors are fully accountable for everything in the manuscript, including any content an AI tool helped produce. If the AI introduces an error, that error is yours.
3. Disclose substantive use
If you used AI to draft or substantially edit text, generate images, or help analyse data, disclose which tool you used and how — usually in the Methods or Acknowledgements, following the journal's policy.
The risks to watch for
- Fabricated citations — AI often invents plausible-looking references that do not exist. Verify every citation against the original source.
- Hallucinated facts — AI can state falsehoods confidently. Fact-check everything, especially clinical or scientific claims.
- Plagiarism — AI can reproduce text closely from its training data. Check originality (see our guide to types of plagiarism).
- Confidentiality — do not paste confidential or unpublished manuscripts (including papers you are reviewing) into public AI tools.
Using AI responsibly — a short checklist
- Never list AI as an author.
- Verify every AI-suggested fact and citation.
- Disclose substantive use as the journal requires.
- Keep confidential material out of public tools.
- Make sure the final work is genuinely yours and accurate.
Key takeaways
- AI can assist, but it cannot be an author and cannot take responsibility.
- Disclose substantive AI use; you remain accountable for the content.
- Verify every citation — fabricated references are the most common failure.
- Protect confidentiality; never upload others' unpublished work.