Can you use AI to write a research paper? What journals allow
Short answer: yes, within limits, at most journals. The major publishing bodies agree on four points: an AI tool cannot be an author, the human authors are responsible for every word, claim and reference, substantial use of generative AI must be disclosed, and AI-generated images are restricted or banned. Language editing with AI is widely accepted; letting a model invent text, data or citations you have not checked is not. Always read the current author guidelines of the journal you are submitting to — policies are still being revised.
What the main policies say
- COPE (Committee on Publication Ethics): AI tools cannot meet the requirements of authorship because they cannot take responsibility for the work; authors who use them must say how, and remain fully responsible for the content.
- ICMJE (medical journals): chatbots should not be listed as authors; authors should disclose AI assistance at submission, in the cover letter and in the manuscript, and must ensure there is no plagiarism in AI-produced text.
- Springer Nature: large language models do not satisfy authorship criteria; their use should be documented, typically in the Methods section. AI-assisted copy editing does not need to be declared.
- Elsevier: generative AI may be used to improve readability and language under human oversight, and its use is declared in a dedicated statement at submission; AI tools may not be listed or cited as authors.
- IEEE: content generated by AI — text, figures, images or code — is disclosed in the acknowledgments, naming the system and where it was used; using AI only for editing and grammar is generally outside the policy, though disclosure is still recommended.
Usually acceptable
- Correcting grammar, spelling and readability of text you wrote.
- Translating your own draft into English, followed by your own careful review.
- Brainstorming an outline or alternative wordings that you then rewrite.
- Summarising papers to decide what to read — and then reading them.
- Help with code, provided you check it and describe it as the journal asks.
Usually not acceptable
- Listing an AI tool as an author or co-author.
- Submitting generated text, results or interpretation as your own work without disclosure.
- Citing references a model produced without checking that each one exists and says what you claim — language models invent plausible references.
- Generating or altering research images, figures or data with AI, unless that is part of the stated method.
- Uploading a manuscript you are peer reviewing to an AI tool: most publishers forbid it because the manuscript is confidential.
How to disclose AI use
Say which tool and version you used, what you used it for, and which sections it affected, and confirm that the authors reviewed and take responsibility for the result. Put it where the journal asks — the methods, the acknowledgments or a separate declaration. For example:
“During the preparation of this work the authors used ChatGPT (GPT-4o) to improve the language of the Introduction and Discussion. The authors reviewed and edited the output and take full responsibility for the content of the publication.”
Checks to run before you submit
- Verify every reference against the source: open each DOI and confirm the title, authors and the claim you cite. The citation generator builds references from the indexed record rather than from memory.
- Run a similarity check, because AI text can echo published wording — see the plagiarism checker.
- Read every sentence as its author. If you cannot explain or defend a sentence, rewrite it.
- Write the disclosure statement while you still remember what the tool did.
An AI detector score is not what editors judge you on, and no detector is reliable enough to prove authorship — see are AI detectors accurate?. Honest disclosure and checked content are what matter.