Graduate schools now widely expect a student to state, somewhere in the thesis or dissertation, whether and how generative AI tools were used in producing it. This is a separate requirement from a journal’s AI-disclosure policy: it is set by the graduate school, the department, or the student’s advisory committee, and it applies to the manuscript as a formal degree-conferring document, not to a submitted article. Journal-level AI disclosure for a manuscript submitted to Nature Portfolio, IEEE, or a journal following the ICMJE Generative AI Policy is a related but distinct requirement (see the Nature Portfolio AI Policy and IEEE Generative AI Policy) — this guide covers the thesis/dissertation-specific version.
Why graduate schools ask for a separate disclosure
A thesis or dissertation is submitted as evidence that the candidate personally did the intellectual work: framed the problem, conducted the research, and produced the writing. Because generative AI tools can now draft prose, summarize literature, suggest analysis, or generate figures, graduate schools have started requiring students to make explicit what a reader otherwise could not tell from the text alone — which parts, if any, involved AI assistance, and of what kind. This sits alongside academic-integrity policy rather than replacing it: a disclosure statement documents assistance that the institution has judged permissible; it does not retroactively make impermissible use acceptable.
Real examples of graduate-school AI-disclosure requirements
Policies are still being written and vary by institution, but several public examples illustrate the pattern:
- Auburn University requires that AI usage by graduate students in research and writing be disclosed and approved by the student’s advisory committee, with AI-use disclosure built into the official thesis/dissertation format requirements.
- Peer-institution policy reviews circulated by graduate schools (for example, a precedent review compiled by the University of Texas at Austin’s provost office) cite institutions such as the University of Louisiana at Lafayette, whose guidance places disclosure prominently in the manuscript — in the introduction, the methods section, or a dedicated GAI disclosure statement following the acknowledgements section — and may require an appendix documenting prompts used and edits made.
- Virginia Commonwealth University‘s guidance, per the same review, draws a line between routine editing assistance and what it calls significant assistance by generative AI, treating the latter as requiring more explicit disclosure.
- Rowan University‘s interim Academic Integrity Policy for Graduate Examination, Thesis, and Dissertation Components requires students to affirm responsibility for the accuracy, originality, and integrity of the final submission — the disclosure obligation is framed as part of that affirmation.
Because there is no single national or international standard for thesis/dissertation AI disclosure comparable to the ICMJE’s guidance for journal manuscripts, the first and most important step for any graduate student is to check the specific requirement set by their own graduate school, department, and advisory committee — policies differ meaningfully even among peer institutions, and a program-level requirement can be stricter than the graduate school’s baseline.
Where the disclosure typically goes
Across the institutional examples above, three placements recur, and some institutions accept more than one:
- A dedicated AI-use disclosure statement, often placed immediately after the acknowledgements section or on its own page near the front matter — the most common pattern where a graduate school has a formal template.
- A note within the Methods (or equivalent) section, where AI use is described alongside other research methods and tools — more common when AI was used for analysis rather than for prose.
- A brief mention in the introduction, where AI-assisted editing of the writing itself is disclosed as a general statement rather than tied to a specific section.
Some institutions additionally expect an appendix listing the specific prompts used and a description of what the student changed or verified afterward, particularly where AI contributed to drafting text or generating a figure. Where a graduate school provides a template or required wording, use it exactly — thesis/dissertation formatting review is often a hard gate before a document can be deposited, and an idiosyncratic disclosure statement risks a formatting rejection independent of its content.
What typically needs disclosing — and what usually does not
Graduate-school policies are still converging, but a consistent distinction shows up across the examples above and across most published frameworks: the more the tool substituted for the student’s own intellectual work, the more disclosure (and often committee approval) is expected.
AI-assisted editing
Using an AI tool to check grammar, suggest wording, or polish sentences the student already wrote is the case most often treated as light-touch — comparable to a grammar checker or a human copyeditor. Some institutions exempt this from disclosure entirely (mirroring how several journal policies, including Nature Portfolio’s, treat routine AI-assisted copyediting of an author’s own text as not requiring a Methods-section disclosure); others still want it named briefly, just without committee pre-approval. Check the specific policy — editing is the category most likely to be treated leniently, but it is not universally exempt.
AI-assisted analysis
Using an AI tool to help design a coding scheme, summarize literature, suggest statistical approaches, or assist with qualitative-data analysis sits in a middle category. This is the kind of use Virginia Commonwealth’s guidance and similar frameworks flag as significant assistance requiring explicit disclosure of the tool, its purpose, and the scope of its application — because it can shape findings or interpretation, not just phrasing. Most policies expect this documented specifically enough that a committee member could understand what the AI tool did and what the student verified independently.
AI-generated content
Text, images, or analysis that an AI tool produced and the student incorporated with minimal alteration is treated most strictly. Many institutions require explicit advisory-committee approval before this kind of use, not just after-the-fact disclosure, and some prohibit it outright for core original-contribution chapters (as distinct from, say, an AI-generated illustrative figure used with permission and disclosed as such). No generative AI tool can be listed as an author or co-author of a thesis or dissertation, consistent with the same authorship-accountability principle underlying the ICMJE, COPE, and journal-publisher positions on AI and authorship.
A practical checklist before submission
- Find your specific graduate school’s AI-use policy (not just a general university-wide AI statement) and your department’s or advisor’s expectations — these can add requirements the graduate school’s baseline does not.
- Identify every point in the research and writing process where an AI tool was used, and sort each use into editing, analysis, or content-generation categories.
- Ask your advisor or committee, before submission, whether any of your AI-assisted analysis or content needed pre-approval rather than only disclosure after the fact.
- Use your institution’s required disclosure template and placement if one exists; if none exists, default to a short, specific statement placed after the acknowledgements, naming the tool(s), the task(s), and what you verified independently.
- Keep records (prompts, tool version, dates) for any substantive AI-assisted analysis or content generation in case your committee or graduate school asks for them during formatting review or the defense.
How this differs from journal-manuscript AI disclosure
A thesis or dissertation chapter that is later adapted into a journal article may need two separate disclosures written for two different audiences: the graduate school’s thesis-format disclosure, and the target journal’s own AI-disclosure requirement at submission (see the CASRAI guide on AI disclosure laws and publisher policy and on disclosing AI assistance in a CRediT statement). The two are not interchangeable — satisfying one does not automatically satisfy the other — so authors converting dissertation chapters into articles should check both. See CASRAI’s guide on publishing a thesis or dissertation as journal articles for more on that conversion process, and the guide on AI tools for dissertation writing for what these tools actually do at each stage.
Frequently asked questions
Do all graduate schools require an AI-disclosure statement in theses and dissertations?
No — requirements vary by institution and are still being introduced or updated. Some graduate schools have a formal, template-driven requirement; others leave it to departmental or advisor discretion; a growing number are adding explicit language as of 2025-2026. Always confirm with your own graduate school’s current thesis/dissertation formatting guidelines rather than assuming a peer institution’s policy applies.
Does using an AI grammar or citation tool always require disclosure?
Not necessarily. Several institutional policies treat routine AI-assisted copyediting of a student’s own already-written text as exempt or minimally disclosed, similar to how some journal AI policies treat basic copyediting differently from substantive drafting or analysis assistance. Confirm the specific line your graduate school draws — editing is not a universally defined category.
Can an AI tool be listed as a co-author or acknowledged as a contributor to a thesis?
An AI tool cannot be listed as an author or co-author, consistent with the accountability-based reasoning that underlies AI-and-authorship guidance from ICMJE, COPE, and major publishers: authorship requires the ability to take responsibility for the work, which a tool cannot do. Naming a tool in an acknowledgements or disclosure statement, describing what it was used for, is a different and generally expected practice.
What happens if a student fails to disclose substantial AI use in a dissertation?
Consequences are set by each institution’s academic-integrity policy, and can range from a formatting-review rejection to an academic-integrity proceeding, depending on the extent of undisclosed use and whether it affected the originality or accuracy of the work. This is treated as an academic-integrity matter, not merely a formatting oversight, at institutions with an explicit disclosure requirement.
Where can I find my own institution’s specific policy?
Check your graduate school’s thesis/dissertation formatting guide (often published alongside style and margin requirements), your department’s graduate handbook, and any university-wide AI-use policy — the three do not always say the same thing, and the most specific, most recently updated one usually governs.
