AI Disclosure Rules: What Journals Ask You to Declare About ChatGPT Use.

Yes, in most journals you now need to disclose ChatGPT use in a research paper whenever the tool did substantive work: drafting or editing your text, producing figures, or helping with analysis. Frontiers, Elsevier, and every journal that follows the ICMJE recommendations require a written statement naming the tool, and COPE's position statement rules out listing AI as an author. The common exception is a basic spelling or grammar checker, which most policies leave undeclared.

Why AI disclosure became a written rule almost overnight

Journals wrote AI rules quickly because authorship, the mechanism they normally use to assign responsibility, does not work for a tool. COPE published its position statement on authorship and AI tools in February 2023, and the major publishers folded similar requirements into their author guidelines soon after.

The logic is short. An author vouches for the accuracy and originality of the work, declares conflicts of interest, and signs the copyright and license agreements. COPE's statement points out that an AI tool can do none of this. It cannot take responsibility for a submitted manuscript, and as a non-legal entity it cannot assert conflicts of interest or hold copyright. So the tools were pushed out of the byline and into a declaration. The people stay accountable for everything, including the machine's output, and the statement records what the machine did.

That history explains a point many first-time authors miss. Most publishers permit AI-assisted language editing. The rules target undeclared use, because an undeclared tool leaves a gap in the chain of responsibility that the whole system is built on.

What does the Frontiers rule actually say?

Frontiers puts the whole requirement in a single sentence, including where the statement goes and what it must contain. The author guidelines for Frontiers in Neurology read:

If generative AI technology was used to produce written or visual content, this must be acknowledged in the acknowledgements section (and methods section if applicable), listing the name, version, model, and source of the AI technology.

Four details, not one. Writing "ChatGPT" alone does not satisfy this rule. You need the name, the version you used, the model that produced the output, and the source, meaning the company or platform behind it. The rule also names two locations. The acknowledgements section carries the statement in every case, and the methods section carries it too when the AI shaped how the research was done rather than only how it was written up.

Is language editing treated the same as content generation?

No. Policies separate three uses, and each carries a different obligation. Language editing calls for a short declaration in most journals, content generation always requires one and is often restricted, and AI-assisted analysis belongs in the methods section.

Elsevier draws the sharpest lines. Its policy on AI in writing permits generative tools only to improve readability and language, requires a declaration when they are used, and states that the policy does not apply to basic spelling and grammar checkers. ICMJE assigns locations by function instead: AI used for writing assistance is described in the acknowledgments, while AI used for data collection, analysis, or figure generation is described in the methods.

The line runs between checking and rewriting. A grammar checker that flags a subject-verb error is a tool in the same category as a dictionary. A chatbot that rewrites your paragraph has produced text, and produced text is what these rules cover. If you paste a section into ChatGPT and paste something back, you are on the disclosure side of the line, even when every idea in the passage is yours.

Authors polishing a second language should not read any of this as discouragement. Journals object to silence, not to editing. What a reviewer means by the English needs improvement is a separate problem with its own fixes, and declared AI help with grammar is a legitimate one in most venues.

Where does the statement go, and what should it say?

Location depends on the publisher, so check the author guidelines rather than copying another paper. Frontiers wants the statement in the acknowledgements, plus the methods if applicable. Elsevier asks for a declaration at the end of the manuscript, above the references. ICMJE journals split it by function, as above.

A statement adapted to the Frontiers rule can be two sentences:

The authors acknowledge the use of ChatGPT (version: May 2026 release; model: GPT-5; source: OpenAI) to edit the language of the Introduction and Discussion sections. The authors reviewed and revised all output and take full responsibility for the content of this manuscript.

Swap in your own tool, version, model, and sections. The template works because it covers all four details Frontiers asks for, and because the second sentence states the thing every policy is after: a person checked the output and answers for it. Keep this statement separate from your CRediT contribution statement, which records what each human author did. The disclosure covers the tool. The contribution statement covers the people.

The AI declaration also joins a long list of statements journals expect on submission day, from ethics approvals to data availability, and the first-submission checklist walks through the rest. Checking a manuscript against the target journal's written rules, including disclosure rules like the Frontiers one above, is also exactly what Waraq does before you submit.

What do editors do with the disclosure, and what about silence?

A disclosed use is handled as routine paperwork. The editor reads the statement, confirms the use sits inside policy, and moves on. The policies quoted above permit language editing outright, so a declared grammar pass gives an editor nothing to act on.

Silence is where the risk lives. The Frontiers rule says "must be acknowledged," and must-language turns an omission into a compliance problem rather than a style preference. An undeclared use that surfaces later, through a reviewer's question or a reader's, is no longer a conversation about your writing process. It becomes a question about the manuscript's integrity, raised at a stage when you never wrote the two sentences that would have answered it.

The asymmetry should settle the decision. The disclosure costs a version number and a line in the acknowledgements. Writing it puts the matter to rest before anyone asks, and leaves the science as the only thing under review.