Your university’s AI policy arrives as a footnote in the assessment brief, then your submission deadline looms, and nobody has explained what an AI use declaration actually looks like. If you used ChatGPT to rephrase a paragraph, Grammarly Go to tidy your prose, or Tesify to structure your methodology chapter, you may be required to disclose it — and getting this wrong is now categorised as academic misconduct at institutions including Cambridge, UCL, and Monash. This guide explains how to write an AI use declaration for your thesis in 2026, with ready-to-paste templates for every use type and citation formats in APA 7 and MLA 9.

What is an AI use declaration for a thesis?

An AI use declaration is a formal, written statement — separate from the body of your thesis — acknowledging that you used one or more AI tools during preparation, and explaining precisely how. It is distinct from your bibliography: you cite AI the same way you cite any source, but the declaration goes further, describing your workflow so that an examiner can evaluate the originality and intellectual ownership of your work.

The concept emerged from academic-integrity frameworks that already required students to disclose paid editing services and statistical consulting. Generative AI extended the same principle: if a tool contributed to your intellectual output in a material way, that contribution must be visible. Most university frameworks in 2026 draw the line at substantive use — generating or substantially restructuring text, producing analyses, creating visuals — while basic traditional spellcheckers typically fall outside the requirement. The grey zone between editing assistance and content generation is exactly where accurate disclosure matters most.

The declaration is not a confession. Universities that permit AI use with disclosure treat it as a mark of methodological transparency, equivalent to disclosing that you used a statistical software package or hired a professional translator. Framing it that way makes writing one less daunting.

Which universities and journals require one in 2026?

The University of Cambridge’s School of Humanities and Social Sciences publishes an official downloadable declaration template requiring students to identify the tool, the prompt they used, the purpose, and any modifications made to the output. Cambridge categorises undisclosed AI in summative assessments as academic misconduct on a par with plagiarism. Oxford follows a comparable “cite like any source” principle, with Russell Group guidance broadly converging on the same standard since 2024.

In Australia, Monash University and the broader Go8 group have adopted disclosure-over-ban frameworks: use of AI in appropriate stages is accepted, but acknowledgement is mandatory. Institutional policies in the United States vary more widely by programme and instructor, though the major academic publishers have aligned: Elsevier requires a “Declaration of Generative AI and AI-assisted technologies in the writing process” immediately above the reference list, while ICMJE — governing most medical and health journals — requires AI disclosure in the Methods section if the tool was used analytically, or in the Acknowledgements if used for writing assistance only.

If your thesis will be submitted for publication after examination, check the target journal’s author guidelines separately. You may need declarations in two formats: one for your institution and one meeting the journal’s own requirements.

What must you disclose? The five core elements

Across the major frameworks — Cambridge, Elsevier, APA, ICMJE — five elements appear consistently in compliant AI use declarations:

1. Tool name and version. “ChatGPT, GPT-4o, OpenAI” or “Tesify v2.1.” Version matters because model behaviour changes between releases, affecting reproducibility assessments.
2. The specific purpose. Did you use the tool to brainstorm an outline, rephrase a paragraph, search literature, code a data-cleaning script, or generate figures? Vague references to “AI assistance” are insufficient at most institutions.
3. Scope. Which section or chapters were affected? “Chapter 3, Methodology” is far more informative than “throughout the thesis.”
4. How you modified the output. “All AI-generated text was rewritten in my own words and cross-checked against primary sources” tells the examiner that you exercised authorial control rather than passed through unedited output.
5. Confirmation of human accountability. A closing sentence affirming that you take full responsibility for the accuracy, originality, and ethical integrity of the submitted work.

Paperpal’s published guidance adds a practical note on analytical AI use: if the tool assisted with data interpretation, declare it in the Methods section rather than front matter only, following the ICMJE distinction between analytical and writing-assistance use.

What does an AI use declaration look like? Templates by use type

Copy the template that matches your primary use case and adapt it before inserting it into your thesis. Most students will need only one; if you used AI in multiple distinct ways, combine the relevant clauses into a single consolidated statement.

Template 1 — Editing and proofreading only

Template 2 — Drafting or structural assistance

Template 3 — Data analysis or literature searching

Choosing the right template also depends on knowing which features of each AI tool count as generative versus traditional assistance. If you used a suite of writing tools, Paperpal, Trinka, and Writefull compared breaks down which features in each platform involve AI generation, making your declaration more precise and more credible.

Where does the declaration go in your thesis?

There is no single universal placement rule, so check your institution’s submission guidelines first. In practice, three placements are standard:

• Front matter (most common for humanities and social sciences). After the abstract and before the table of contents. Cambridge’s SHSS framework uses this placement. It signals transparency upfront before the examiner reads a word of your argument.
• Appendix. Some UK institutions, particularly for STEM theses, request an appendix containing the declaration alongside logs of AI interactions — full prompt and response transcripts. If your university requires a log, export or screenshot your AI sessions throughout writing, not retrospectively.
• Methods section (analytical use). If AI contributed to data collection, coding, or quantitative or qualitative analysis, declare it within the Methods section where the relevant process is described, following ICMJE convention. A brief cross-reference to front matter is acceptable as well.

For AI-assisted results chapters specifically, writing your results chapter with AI includes a workflow that generates auditable, declaration-ready records at each stage — so you are not trying to reconstruct your process from memory at submission time.

How do you cite AI tools in APA 7 and MLA 9?

The declaration and the citation are two separate obligations. The declaration describes your workflow; the in-text citation and reference entry credit the tool when you reproduce or closely paraphrase its output. Both are required if your thesis body incorporates AI-generated material that was not fully rewritten.

APA 7 format

The APA Style blog recommends citing a specific AI response as a retrievable source when a share link exists:

If no shareable link exists — common when using API access or enterprise accounts — describe the session in brackets: [Large language model response; session not retrievable]. APA also recommends appending the full prompt–response exchange as an appendix so readers can assess the content and your analytical choices.

MLA 9 format

MLA treats the AI tool as a container rather than an author. The MLA Style Center’s current guidance formats the citation as follows:

Note that the prompt itself (or a description of it if the prompt is long) serves as the “title” in MLA — you cite the specific AI response, not the tool in general. For engineering and computer science theses where IEEE format applies across the reference list, the IEEE citation style guide for 2026 covers how the standard currently handles AI-tool references.

It is also worth understanding how AI detection tools read your document alongside your declaration. If examiners run automated screening before reviewing your statement, knowing how AI detectors compare across free and paid tiers helps you anticipate what flags might appear — and why a well-drafted declaration pre-empts any misinterpretation of those results.

What happens if you do not declare AI use?

At Cambridge, submitting AI-generated content without acknowledgement in a summative assessment “constitutes academic misconduct.” The consequence framework mirrors that for plagiarism: mark reduction, module failure, or — for persistent or egregious cases — suspension or withdrawal of the degree. Most Russell Group institutions have converged on equivalent policies since the 2023–24 academic year.

The risk extends beyond the grade. If your thesis is later published and undisclosed AI use is identified by a journal editor — Elsevier, for example, actively checks author declarations against submitted manuscripts — you may face post-publication retraction. For a PhD thesis that forms the basis of a research career, retraction is a disproportionate consequence for something a two-paragraph declaration would have prevented. Disclosure is increasingly treated as a marker of methodological rigour, not a stigma to avoid.

If you are uncertain whether your institution permits the AI use you have already conducted, read what is and is not permitted when using AI in your thesis before your submission date. If you used AI for literature synthesis, using AI to write your literature review faster includes a section on where the integrity boundaries sit for each stage of the review process.

Frequently asked questions