CorrectMyPaperBefore you submit, you will probably ask one practical question: if an AI tool helped me polish this manuscript, do I have to say so? The three largest commercial publishers, Elsevier, Springer Nature, and Wiley, have each published an answer. They agree on more than they differ, but the details decide whether you add a statement to your paper or leave it out, and a wrong guess in either direction is avoidable.
This guide summarises each policy as it reads at the time of writing (October 2026). Publishers revise these pages, and individual journals can add their own rules, so treat this as a map and confirm against your target journal's Instructions for Authors before you submit.
What all three agree on
Three principles appear in every one of the policies:
- AI cannot be an author. Authorship carries accountability for the work, and a tool cannot take it. Springer Nature says large language models do not meet its authorship criteria for exactly this reason, and Elsevier says authors should not list or cite AI tools as an author or co-author.
- You are responsible for everything in the paper. If a tool introduces an error, a wrong claim, or a reference that does not exist, the author answers for it.
- Routine language editing is the low-risk end. Improving grammar, spelling, and readability is treated differently from using AI to generate content, analyse data, or draw conclusions.
Elsevier
Elsevier allows generative AI and AI-assisted technologies in the writing process before submission, but only to improve the language and readability of the work, and with appropriate disclosure. It does not accept AI as a substitute for the author's own critical thinking, expertise, and evaluation.
Two details matter in practice:
- The declaration. When the policy applies, the statement goes at the end of the manuscript, immediately above the references, under the heading "Declaration of Generative AI and AI-assisted technologies in the writing process."
- The exemption. Elsevier states that basic checks of grammar, spelling, and punctuation need no declaration, and that its policy does not relate to tools such as spelling or grammar checkers.
Elsevier also does not allow generative AI to create or alter images in submitted manuscripts, apart from narrow exceptions where the tool is part of the research method.
Springer Nature
Springer Nature's guidance is that using an AI tool for "AI assisted copy editing" does not need to be declared. It defines this as AI-assisted improvements to human-generated text for readability and style, and to ensure that the text is free of errors in grammar, spelling, punctuation, and tone, with the human author remaining responsible.
Where AI goes beyond that, the rule changes. Use of a large language model should be documented in the Methods section (or another suitable section if the paper has no Methods). Springer Nature journals also apply specific rules to AI-generated images and to AI used in peer review, which is outside the scope of this guide.
Wiley
Wiley welcomes thoughtful use of AI and lists improving grammar, spelling, punctuation, tone, and overall readability as acceptable. Like the other two, it excludes tools used for spelling, grammar, and general editing from its disclosure requirement.
Beyond that, Wiley is the most specific about where disclosure belongs:
- Acknowledgments when AI assisted with manuscript drafting, editing, translation, or formatting.
- Methods when AI was used in the research itself: study design, data collection, literature review, code development, or analysis.
Wiley asks authors to name the tool, state its purpose, say whether it affected key arguments or conclusions, and describe how they reviewed and verified the output.
Side by side
| Elsevier | Springer Nature | Wiley | |
|---|---|---|---|
| AI as author | Not allowed | Not allowed | Not allowed |
| Basic grammar and spelling help | No declaration needed | No declaration needed ("AI assisted copy editing") | No declaration needed |
| Beyond basic editing | Declaration statement before the references | Document in Methods | Acknowledgments (editing, drafting) or Methods (research use) |
| Author accountable for content | Yes | Yes | Yes |
The grey area: when editing becomes something else
The exemptions cover tools that improve text you already wrote. They do not obviously cover a tool that writes for you. Fixing a grammatical error is editing. Asking a tool to tighten a paragraph you wrote is still revision of your own text. Asking it to "expand on this point" or "suggest more implications for the discussion" asks the model for intellectual content that was not in your draft. That is the point where a disclosure is expected, and where a journal may ask harder questions.
A workable rule: if every idea, claim, and number in the passage came from you, you are editing. If the tool contributed an idea or a claim, you are generating, and you should disclose it.
Because the exemptions are written around grammar, spelling, and readability, a tool that returns many structural or substantive suggestions sits closer to the line than a spell checker does. If you are unsure which side your use falls on, disclose. A short, accurate statement costs you one sentence. An undisclosed use that an editor later questions costs far more. For the related question of how editors treat detector scores, see AI Detection and Academic Writing: What's Safe, What's Risky, and for the ethics of AI editing generally, Is It Plagiarism to Use AI to Edit Your Research Paper?.
A practical checklist before you submit
- Read the journal's Instructions for Authors. Individual journals can add requirements on top of the publisher's policy, including where the statement goes.
- Decide which side of the line each use falls on. Grammar, spelling, and readability are exempt at all three publishers. Drafting, translation, restructuring, and anything touching your analysis are not.
- Check every suggestion yourself. You are accountable for the final text. Accept the changes you agree with and reject the rest.
- Verify every reference by hand. AI tools can produce citations that look real but do not exist, and that is an integrity problem on its own.
- Write the statement now if you need one. Be specific: the tool, what you used it for, and that you reviewed the output. For example: "The authors used [tool name] to improve the language and clarity of the manuscript. All suggestions were reviewed and verified by the authors, who take full responsibility for the content."
- Keep your original draft. If anyone asks what the tool changed, you can show them.
Where a revision tool fits
A pre-submission review tool works on your existing manuscript: it points out problems and suggests corrections that you accept, dismiss, or edit yourself. That keeps you as the author and the decision-maker, which is what all three policies ask for. You can see what this looks like on a real manuscript with the sample review, then decide for yourself which of the suggestions are plain language fixes and which change substance.
The bottom line
All three publishers draw the same line in roughly the same place. Using AI to improve the language of your own text is accepted, and basic grammar and copy editing needs no declaration. Using it to generate content, analyse data, or draw conclusions must be disclosed, with the statement placed where that publisher expects it. Anything unclear is better disclosed than left out. Check your journal's current wording, keep your own judgement at the centre of every change, and you will be within the spirit of all three policies.