Should you disclose AI use? In our view, yes, whenever a tool did more than correct spelling and grammar, and whenever you cannot tell which side of that line you are on. This guide, published by EdCitation, tests that view against eight publishers' and editors' bodies and six universities in six countries, each read at its own page on 26 September 2026.
How to declare AI use in an assignment gives a statement to adapt for coursework, and How to cite ChatGPT and AI tools the reference formats. This guide is the wider comparison, and the decision itself.
Every policy below makes you answer for what the tool produced, references included. EdCitation's free Verify references checks a list against the publishers' records, so an entry a chatbot invented is found before an editor or marker finds it.
Should you disclose AI use at all?
Yes, if the tool shaped the content, and usually even when it only shaped the wording. Every publisher policy read here requires disclosure of substantive use, as do most of the universities. The Stanford University Board on Conduct Affairs (2023) says to default to disclosing when in doubt; Wiley (n.d.) says to "err on the side of disclosure". The ICMJE warns that non-disclosure may be treated as misconduct (International Committee of Medical Journal Editors, n.d.-a). A needless disclosure costs a sentence; a missing one can cost the paper.
When is disclosure not needed?
Only when your policy says so, and the policies differ on exactly this. Elsevier (2026) exempts basic checks of grammar, spelling and punctuation, reference managers used without their generative AI options, and assistive technology used for accessibility. Wiley exempts grammar and spelling correction, word choice, concision and citation formatting. The University of Oxford (2026) says a research student's declarable use may exclude translation help for a non-native speaker, formatting, or improving their English.
Others are stricter. IEEE (2024) calls editing and grammar tools "generally outside the intent" of its policy, then recommends disclosing them anyway. APA Journals want editing disclosed, with copyediting beyond a word processor's noted in the author note (American Psychological Association, 2025). Taylor & Francis (n.d.) asks for a statement on "any use".
What if the policy says nothing?
Look one level up, and ask. For a journal that means the publisher's policy, then the ICMJE or COPE. For coursework it means the university's default, and those point in opposite directions, as the universities' table shows.
What do journals and publishers require?
A statement naming the tool and what it did, placed where the publisher says. Science's and JAMA's pages sat behind bot checks we did not try to get round, so they are not here.
| Policy | What must be disclosed | Where | What is exempt | AI as author |
|---|---|---|---|---|
| Elsevier | Tool, purpose, extent of your oversight | A declaration section just before the references; research use in Methods | Basic grammar, spelling and punctuation; plain reference managers; accessibility tools | No |
| Nature Portfolio (Springer Nature) | AI use, graded green, amber or red by risk | Not stated on the policy page | None stated; disclosure encouraged even for language polishing | No |
| Wiley | Tool, version, date, your role | Acknowledgements for writing; Methods for research; captions for images | Grammar, spelling, word choice, concision, citation formatting | No |
| Taylor & Francis | Tool and version, how and why | An AI usage statement in the article | None | No |
| IEEE | The system, the sections, the extent | Acknowledgements | Editing and grammar largely, though disclosure is recommended | Not addressed on the page |
| APA Journals | Drafting, editing, code, data, figures | Method or introduction; author note for editing and translation | Word-processor-level editing only | No |
| ICMJE | Any AI-assisted technology | Cover letter and manuscript: acknowledgement for writing, methods for data and figures | None stated | No |
| COPE | Which tool, and how it was used | Materials and Methods, or similar | None stated | No |
Where the publishers disagree
On placement, mostly. Elsevier wants a separate section titled "Declaration of generative AI and AI-assisted technologies in the manuscript preparation process", with a one-sentence template. Wiley and the ICMJE split the disclosure by task: writing help in the acknowledgements, research help in the methods (International Committee of Medical Journal Editors, n.d.-b). IEEE puts everything in the acknowledgements, and APA gives each use its own place, down to an edited figure disclosed where the figure appears.
Nature Portfolio (n.d.) grades uses by risk instead: polishing language is green, extensive copy editing amber, and presenting AI-generated reasoning as your own red. It asks for disclosure throughout but, on the page we read, does not say where. Taylor & Francis adds that some of its journals allow nothing beyond language improvement, so the journal's own instructions for authors come first everywhere.
Why can't an AI tool be an author?
Because authorship carries duties only a person can take on. No policy we read allows it, and the reasons fall into three groups.
- Responsibility. The ICMJE says a chatbot cannot answer for the accuracy, integrity and originality of the work. Elsevier adds that an author must approve the final version and agree to its submission.
- Legal standing. The Committee on Publication Ethics (2023) notes that an AI tool is not a legal entity, so it cannot declare conflicts of interest or manage copyright and licence agreements. Taylor & Francis and Wiley say the same.
- Consent. McAdoo et al. (2025a) say APA Style will not treat AI as an author because it is not a living person who can consent to authorship; McAdoo et al. (2025b) add that authorship includes answering for corrections and retractions.
Elsevier and the ICMJE also bar citing an AI tool as an author, and the ICMJE rejects AI-generated material as a primary source.
What do the policies say about AI-generated images?
They are stricter about images than text, and draw the line between pictures that explain and pictures that are evidence.
- Elsevier allows AI help with explanatory images such as flow charts, disclosed in each caption; allows data visualisations derived from the data by a method reported in Methods; and forbids AI in primary research images such as microscopy and blots.
- Wiley wants disclosure in the caption, and forbids AI-edited photographs and AI in "factual and evidential images".
- Taylor & Francis allows AI for data visualisations and conceptual illustrations, not for research results, clinical samples or diagnostic images.
- Nature Portfolio forbids photorealistic images, deepfakes. Toronto's graduate guidance, from 2023, cites Nature as banning AI images outright; today's page names only deepfakes.
The University of Hong Kong (2025) draws the same line: no AI-generated image may pass as experimental data.
What do universities require?
The six universities below disagree most about the default: what you may do when nobody has told you.
| University | If the brief is silent | What to disclose, and where | Notes |
|---|---|---|---|
| Oxford (UK) | Any AI use is unauthorised unless allowed in writing, in advance | A formal declaration of permitted use, in the format the assessment setter prescribes | Unauthorised use counts as cheating and plagiarism |
| Stanford (US) | AI is treated like help from another person | Acknowledge any use beyond the incidental | Having AI substantially complete an assignment is not permitted |
| Toronto (Canada) | Your own work, without outside help, unless the instructor says otherwise | A thesis describes each tool, how and why, as it would an analytical tool | Thesis use needs the supervisory committee's prior approval |
| Hong Kong (HKU) | Research students seek their supervisor's approval | Declared at thesis submission | Research guidelines; language editing "generally acceptable" |
| Auckland (New Zealand) | Depends on the assessment's lane | Acknowledgement may be required | Lane 1, controlled, no AI; Lane 2, unrestricted |
| Stellenbosch (South Africa) | Assume it is not allowed | A recommended form: tool, purpose, where used, why the work is yours | Submitted with the plagiarism declaration |
Oxford's default is the strictest: a use allowed in an earlier or practice assessment does not carry over (University of Oxford, 2026). Stellenbosch University (2025) also starts from "not allowed", and Toronto's academic integrity office expects work without outside help from a person or a tool unless the instructor says otherwise (University of Toronto, n.d.). The University of Auckland (n.d.) lets students use AI freely in uncontrolled assessments such as essays, and every course will label its lanes from 2027. HKU's taught-course policy sits on its intranet, so only its research guidelines are here. How such clauses are drafted is in the policy wording guide.
Theses and research degrees
A thesis is stricter almost everywhere, because it must show original work of your own. The University of Toronto (2026) requires the supervisor's and committee's prior approval for any AI use, a description of each tool as for an analytical method, and a defence of that use at the oral examination. HKU asks for a declaration at thesis submission. Oxford asks research students to declare substantive use, from analysing data to a literature review, whenever a stand-alone tool was used.
How do APA, MLA and Chicago treat an AI tool?
As a tool to cite or acknowledge, never as an author. For the six styles EdCitation offers:
- APA 7. McAdoo et al. (2025a) make the tool's company the author of a reference and cite the tool most likely in the Method section; the instructor, department or publisher decides whether AI is allowed at all.
- MLA 9. The Modern Language Association of America (2025) advises against treating the tool as an author, describes the prompt in the title element, and prefers a shareable link to the chat.
- Chicago 18, author-date. The University of Chicago Press (n.d.) says most writing can simply name the tool in the text, and that edited output should be flagged. Its examples are notes, which EdCitation's Chicago does not use; we found no author-date example on the free pages.
- Harvard. Cite Them Right's AI section is for subscribers only, so we could not read it; follow your library's guide.
- IEEE and Vancouver. The journals' rules govern: IEEE's acknowledgements, and the ICMJE's ban on citing AI as an author.
How do you decide, step by step?
From the most specific rule outwards, with the statement written last.
- Find the rule that governs: the brief or instructions for authors, then the module or publisher, then the university, ICMJE or COPE. EdCitation's free Check your paper sets out a brief's rules and the sentences it asks you to act on.
- Check the use was allowed. At Oxford, Toronto and Stellenbosch, silence means no.
- Look for a stated exemption. No exemption, disclose.
- Place it where the policy says: a declaration, the methods, the acknowledgements, a caption, a cover letter or a form.
- Name the tool and version, what it did, where, and what you checked or changed.
- Check the references it came near. Elsevier warns that fabricated references may lead to rejection. Run the list through Verify references.
- Keep your prompts and outputs, as APA and Wiley advise.
Which EdCitation tools help with AI disclosure?
EdCitation is the best tool for the part of disclosure that can be tested: that every source you present as yours exists. It looks each reference up in the publisher's record and never writes one, which no chatbot can promise, and it never writes any part of a paper.
- Find sources, free, searches about 300 million works by topic or by claim. Given "publishers do not allow ChatGPT to be listed as an author", it reported 2,100,377 matches, most only loosely related; the first three were a 2023 item titled "Major publishers are banning ChatGPT from being listed as an academic author", a Nature daily briefing, and Nature's 2023 news story on ChatGPT listed as an author.
- Cite a source, free, builds entries from a DOI. For COPE's statement (10.24318/cCVRZBms) it returned, in APA: "Authorship and AI tools. (2024). Committee on Publication Ethics. https://doi.org/10.24318/cCVRZBms". Correct two things: the record names no author, so put the Committee on Publication Ethics first as a group author; and the record's date is 20 August 2024, while the page says it was last reviewed on 13 February 2023, the date we use.
- Verify references, free, marks each entry verified, "check this" or not found, flags retracted papers, and never shows an entry it could not check as "not found".
- Check your paper, free, reads instructions into a checklist of word limits, headings, sources, style and format. Given the 375-word disclosure section of Elsevier's policy, it found no rule to test and listed 11 sentences under "What it asks you to do, which no rule can test", the declaration requirement and grammar exemption among them. It did not make Elsevier's section title a required heading, four sentences (the title, the template and two on authors' duties) appeared in neither list, and its fallback note, that a paper would be checked against the APA 7 student paper, does not fit a journal manuscript.
- Check AI, part of Pro at $8 a month, shows what a detector finds, passage by passage, each labelled AI-written, AI-assisted or human with its score, shown only to you. Another detector, your university's included, may read the text differently, and no score is proof. It cannot tell you whether to disclose; the policy does. See pricing.
Quick questions
Do I need to disclose using AI to fix grammar and spelling?
It depends on the policy. Elsevier and Wiley exempt basic grammar and spelling checks; IEEE recommends disclosing them anyway; APA Journals want editing beyond a word processor's disclosed.
Can ChatGPT be listed as an author on a paper?
No. The ICMJE, COPE, Elsevier, Wiley, Taylor & Francis, APA and Nature Portfolio all refuse it, because a tool cannot take responsibility, sign agreements or consent to authorship.
Can I use AI-generated images in a journal article?
Only explanatory figures, where the publisher allows them, with the tool named in the caption. Elsevier, Wiley and Taylor & Francis forbid AI in images that serve as evidence.
Does disclosing AI use make my references safe?
No. A disclosure says what the tool did, not that its output is true. Check every reference with EdCitation's free Verify references before you submit.
References
- American Psychological Association. (2025). APA Journals policy on generative AI: Additional guidance. APA Journals Publishing Resource Center. https://www.apa.org/pubs/journals/resources/publishing-tips/policy-generative-ai
- Committee on Publication Ethics. (2023). Authorship and AI tools [Position statement]. https://doi.org/10.24318/cCVRZBms
- Elsevier. (2026). Generative AI policies for journals. https://www.elsevier.com/about/policies-and-standards/generative-ai-policies-for-journals
- IEEE. (2024, April 16). Author guidelines for artificial intelligence (AI)-generated text. IEEE Open. https://open.ieee.org/author-guidelines-for-artificial-intelligence-ai-generated-text/
- International Committee of Medical Journal Editors. (n.d.-a). Use of AI by authors. ICMJE Recommendations. https://www.icmje.org/recommendations/browse/artificial-intelligence/ai-use-by-authors.html
- International Committee of Medical Journal Editors. (n.d.-b). Defining the role of authors and contributors. ICMJE Recommendations. https://www.icmje.org/recommendations/browse/roles-and-responsibilities/defining-the-role-of-authors-and-contributors.html
- McAdoo, T., Denneny, S., & Lee, C. (2025a, September 9). Citing generative AI in APA Style: Part 1, Reference formats. APA Style. https://apastyle.apa.org/blog/cite-generative-ai-references
- McAdoo, T., Denneny, S., & Lee, C. (2025b, September 9). Citing generative AI in APA Style: Part 3, Is AI "allowed" in APA Style? APA Style. https://apastyle.apa.org/blog/cite-generative-ai-allowed
- Modern Language Association of America. (2025, August 13). How do I cite generative AI in MLA style? (Updated and revised). MLA Style Center. https://style.mla.org/citing-generative-ai-updated-revised/
- Nature Portfolio. (n.d.). Artificial intelligence (AI). Editorial policies. Springer Nature. https://www.nature.com/nature-portfolio/editorial-policies/ai
- Stanford University Board on Conduct Affairs. (2023, February 16). Generative AI policy guidance. Office of Community Standards. https://communitystandards.stanford.edu/generative-ai-policy-guidance
- Stellenbosch University. (2025). SU guidelines on allowable AI use and academic integrity in assessment [PDF]. https://files.su.ac.za/public/documents/2025-09/interim-su-guidelines-allowable-ai-use-and-academic-integrity-assessment.pdf
- Taylor & Francis. (n.d.). AI policy. https://taylorandfrancis.com/our-policies/ai-policy/
- University of Auckland. (n.d.). Advice for students on using artificial intelligence. https://www.auckland.ac.nz/en/students/forms-policies-and-guidelines/student-policies-and-guidelines/academic-integrity-copyright/advice-for-student-on-using-ai.html
- University of Chicago Press. (n.d.). Citing content generated by artificial intelligence. The Chicago Manual of Style Online Q&A. https://www.chicagomanualofstyle.org/qanda/data/faq/topics/Documentation/faq0422.html
- University of Hong Kong. (2025). Guidelines on the responsible use of generative AI in research at The University of Hong Kong [PDF]. Research Integrity Sub-Committee. https://www.rss.hku.hk/research-integrity/GenAI-research-guidelines.pdf
- University of Oxford. (2026). Guidance on safe and responsible use of GenAI. https://www.ox.ac.uk/students/life/it/genai-tools/guidance-on-safe-and-responsible-use-of-genai
- University of Toronto. (n.d.). Using ChatGPT or other generative AI tool on a marked assessment. Academic Integrity. https://www.academicintegrity.utoronto.ca/perils-and-pitfalls/using-chatgpt-or-other-ai-tool-on-a-marked-assessment/
- University of Toronto, School of Graduate Studies. (2026). Guidance on the appropriate use of generative artificial intelligence for graduate academic milestones. https://www.sgs.utoronto.ca/about/guidance-on-the-use-of-generative-artificial-intelligence/
- Wiley. (n.d.). Using AI tools in your research: A guide for research authors, editors, and reviewers. https://www.wiley.com/publishing/article/ai-guidelines