
Sample AI Use Declaration for Your Dissertation
You wrote the dissertation. But what should you say if ChatGPT helped with the outline, Grammarly suggested changes, or a human proofreader checked the final draft?
You’ve finished your dissertation. Then, just before submission, you find an unexpected requirement: declare your use of artificial intelligence.
Does Grammarly count? What if ChatGPT helped you find better search terms but never wrote a paragraph? Should you mention the professional proofreader who corrected your grammar? And what if an AI tool rewrote a passage you originally drafted yourself?
There’s another complication students may not have considered: some generative AI systems now place invisible statistical watermarks in supported text. These don’t automatically establish misconduct, but they make it especially important to be precise about whether AI generated wording or simply offered feedback.
This guide gives you five ready-to-adapt declarations, explains when each is appropriate and shows how to describe human proofreading without incorrectly labelling it AI assistance. The examples are starting points, not substitutes for your university’s mandatory form.
Do You Need an AI Use Declaration for Your Dissertation?
It depends on your university and the specific assessment. Some institutions require a declaration for any permitted AI use; others set thresholds, exclude particular routine tools, ask for an acknowledgement or require students to complete a prescribed submission statement. Rules can differ between taught dissertations and doctoral research, even at the same university.
For example, the University of Oxford’s summative-assessment policy requires assessment setters to specify the permitted uses of AI and the declaration format, and requires students to follow those instructions. UCL’s three assessment categories help staff communicate whether AI is prohibited, assistive or integral, but the assessment setter determines the conditions for the task.
That means an example copied from a website cannot override a course handbook, submission portal or supervisor’s written instructions. Where your institution supplies a template, use that template first.
What to Include in Your Declaration
A useful AI use statement normally answers four questions. Your institution may ask for additional information, such as dates, prompts, outputs, affected sections or an appendix.
Describe the assistance, not just the tool
- Which tool? Name the specific service and, if relevant, the feature used: ChatGPT, Grammarly’s grammar suggestions, Elicit literature search or another tool.
- What did it do? Say whether it suggested search terms, explained concepts, reviewed grammar, produced text or rewrote passages.
- Where and how much? Identify the stage or sections affected and the extent of generated wording where your policy requires it.
- What did you do? Explain what you independently verified, reviewed, rejected or wrote yourself—without making claims you cannot substantiate.
Keep wording factual. Instead of saying “AI was used for assistance,” write “I used ChatGPT to suggest alternative database search terms during the literature-search stage; I selected and verified sources independently.” Instead of saying “AI checked my thesis,” distinguish identifying possible errors from generating replacement sentences.
Do not routinely add “No AI tool generated any words in this dissertation” if you accepted AI-generated rewrites or replacement sentences. Even with light language corrections, describe the feature and extent of assistance rather than using a formula that might be too absolute.
Five Sample AI Use Declarations to Adapt
Choose the scenario closest to your actual workflow. Replace every bracketed placeholder, delete tools you did not use and add any assistance that the university requires you to disclose. Only use a declaration if it is true.
When a writing assistant corrected language in your own draft
AI Use Declaration
I used [tool and feature, e.g. Grammarly grammar and spelling suggestions] to identify possible spelling, punctuation and grammatical errors in material that I had drafted myself. I reviewed the suggested corrections and accepted or rejected them individually. The tool was not used to develop my research question, interpret my data or create my academic arguments. [If any passages were substantially rewritten, describe that separately.]
[Name] · [Date, if required]
Use this only if it accurately reflects the features you used. Some writing assistants offer both ordinary correction and generative rewriting; those are not necessarily treated the same way by your university.
When you used research tools but wrote the dissertation yourself
AI Use Declaration
I used [Elicit/another research tool] during the literature-search stage to identify potentially relevant publications and organise initial research leads. I consulted the original publications and independently assessed the evidence before citing them. I also used [grammar-checking tool] to identify possible language errors in my own drafts, reviewing suggested corrections before applying them. The research design, analysis and conclusions were developed by me. [State any generated text used, if applicable.]
[Name] · [Date, if required]
Do not assert that you read every paper in full if you did not. Explain your verification process accurately.
When AI assisted your thinking and a person checked the language
AI Use Declaration
I used ChatGPT during the planning stage to generate alternative search terms and questions that helped me assess the scope of my dissertation. I evaluated the suggestions independently and did not use ChatGPT to produce dissertation chapters or interpret my findings. My completed draft was subsequently proofread by [Vappingo/name of proofreader] for language and presentation, using visible corrections and comments. This human proofreading is disclosed here as third-party assistance [delete or move to a separate acknowledgement if your institution requires AI assistance and human proofreading to be reported separately]. The research, analysis and academic conclusions remain my own.
[Name] · [Date, if required]
Use the human-proofreading sentence only if that service was actually used and the level of proofreading was authorised.
When several tools helped with planning, comprehension and feedback
AI Use Declaration
In preparing this dissertation, I used [tools] for the following purposes: [research tool] to identify and screen possible sources; ChatGPT to clarify methodological terminology and suggest questions for me to investigate; and [language tool] to flag grammar and consistency issues in my own writing. I checked factual information and source material independently. I made the final decisions about the research question, methods, argument, interpretation and conclusions. The AI tools were not used to generate substantive dissertation sections. [Specify any AI-generated wording retained in the final document and any additional uses, if applicable.]
[Name] · [Date, if required]
“Extensive assistance” is not automatically permitted. This sample is only suitable if your assessment explicitly allows the described activities.
When you genuinely did not use the tools covered by your university’s declaration requirement
AI Use Declaration
I confirm that I did not use generative AI or other AI-assisted tools covered by [university/assessment declaration policy] in the preparation of this dissertation. The research, analysis and submitted writing are my own work. [Identify any separately declarable human proofreading or collaboration, if required.]
[Name] · [Date, if required]
Do not use this if you used ChatGPT, an AI research assistant or a generative rewriting feature that falls within the policy. If your university excludes ordinary spell-checking, follow its definition rather than claiming no software assistance of any kind.
These templates are deliberately more precise than generic declarations that claim every submitted word was produced without machine assistance. Modern writing tools contain different features, and the actual role of the tool is what needs to be recorded.
What If AI Rewrote Part of Your Dissertation?
This is the scenario where a standard “grammar checking only” declaration may become misleading. Asking ChatGPT to identify unclear sentences is different from asking it to rewrite three pages of your literature review. The second request generates new wording, even if the original ideas and sources came from you.
If your assessment permits generative rewriting, describe it honestly and at the level of detail your institution requires. A statement might say:
“I used [AI tool] to suggest alternative wording for [identified section(s)] during the revision stage. I reviewed and amended those suggestions against my original draft and sources. [Describe how much generated wording was retained and any required prompt records.] This use was permitted under [relevant assessment guidance].”
If your assessment does not permit generated prose or substantial rewriting, a declaration does not retrospectively authorise it. Preserve your genuine drafting record and seek guidance through your supervisor, module lead or student-support service before submission. Do not try to disguise generated text through repeated paraphrasing.
For the broader distinction between feedback, light editing and generated prose, read Can I Use ChatGPT for My Dissertation?.
Could AI Editing Leave an Invisible Text Watermark?
Potentially, yes, where a supported model generates enough new wording. A statistical text watermark is not a hidden Word character or a tag attached to the document. It is a pattern introduced into a model’s choices of words or tokens. Some patterns can survive copying and pasting, although detection is imperfect.
In October 2026, OpenAI announced textGrain for eligible ChatGPT and Codex text output in the European Union, with opt-in availability for supported API models globally. OpenAI says a watermark may indicate that its model generated or processed wording but does not establish how much the human contributed, who used the tool or whether disclosure was required. Its text detector is not universally available.
Anthropic’s explanation of Claude text watermarking makes an especially useful distinction for students: if Claude only fixes a few grammatical errors, there may be too little new wording for a detectable signal. Heavier rewriting gives the model more word choices and potentially more room for a watermark. Google DeepMind also documents SynthID text watermarking for supported Gemini-generated text.
A watermark is not a misconduct verdict. Nor does an undetected watermark prove that the writing is human. What matters for a declaration is what you actually asked the system to do and what generated wording you used. A conventional AI-detector percentage and a provider-specific watermark check are also different types of evidence; see AI Watermarks vs. AI Detectors.
Do You Need to Declare Human Proofreading?
Human proofreading is not generative AI use. However, some universities require students to acknowledge third-party assistance separately or alongside AI use, and some restrict what a human proofreader is allowed to change. Do not assume it is universally permitted or that anything an editor does counts as acceptable proofreading.
The University of Cambridge’s proofreading guidance, for example, distinguishes identifying language and presentation problems from altering academic content, arguments and structure. It emphasises that students retain responsibility for their work and that tracked changes and comments help keep intervention visible.
Where permitted, a human proofreader may help identify typographical problems, language inconsistencies, awkward phrasing and places needing author clarification. Academic decisions—such as which sample size is correct, whether a causal claim is supported or how findings should be interpreted—must remain with the researcher.
If the university only requests AI use, put human proofreading in a separate acknowledgement if required. If it requests all external help in one statement, make the distinction explicit:
“The completed dissertation was reviewed by a professional human proofreader at Vappingo for [the permitted tasks actually completed]. Proposed corrections and queries were returned using Track Changes and comments. I reviewed the changes and retained responsibility for the research, arguments, data and final submission.”
What a Certificate of Human Editing Can Document
The original version of this article described a Vappingo Certificate of Human Editing as conclusive reassurance about the origin of a dissertation. That would go too far. A certificate can document a human editorial service; it cannot prove who originally wrote every sentence of the manuscript.
Where Vappingo supplies a certificate for an order, it may help document the identity of the editing provider, the nature of the service and its scope, together with any dates or editor details stated on the actual certificate. The edited file’s Track Changes and comments can also show what the proofreader proposed. Those records can be useful when a university asks about third-party assistance.
They do not establish that no AI was used before the file was submitted for editing, remove a watermark from earlier AI-generated text, guarantee a detector will not return a false positive or override a university rule. Students should confirm with Vappingo what documentation is available for their order, and with their institution whether it may be submitted.
For additional context on the limits of AI-authorship evidence, see Can Universities Actually Prove You Used AI?.
Declaration Mistakes to Avoid
Claiming every word is human-written
Avoid an absolute statement if you accepted AI-generated replacement wording. Explain the actual feature and level of assistance instead.
Calling all AI use “research”
Finding search terms, summarising papers, creating an outline and drafting paragraphs are distinct activities with different academic implications.
Using somebody else’s template unchanged
Delete unused tools and adapt every example. A polished but inaccurate declaration is worse than a plain factual statement.
Assuming human proofreading needs no permission
Check your university’s third-party assistance rules and ask for a permitted, clearly defined level of language intervention.
Other common errors include treating a required declaration as optional; failing to disclose AI use because it seems minor; asserting every suggested source was verified when it was not; and assuming a permitted AI tool means every feature within that tool is also permitted.
There is no need to overstate your diligence either. If you used a tool to search for studies and checked only the sources you cited, say that. If ChatGPT generated a draft you did not use, record that where policy requires it, without pretending the output formed part of the final manuscript.
Your Final Submission Checklist
Check the declaration against the real workflow
- Read the dissertation-specific AI and proofreading instructions, not just a general university policy.
- Use the exact declaration form and placement specified by the department.
- Name the tools and functions actually used, including generative rewriting where relevant.
- Check whether human proofreading belongs in a separate acknowledgement of third-party assistance.
- Remove absolute claims that are not strictly accurate.
- Keep genuine drafts, notes and permitted records of AI use where appropriate.
- Confirm that any supporting certificate or editing record is accepted by the submission process before attaching it.
If you are considering professional proofreading, it is worth showing your proposed editing brief and your university’s policy to the proofreader before the work begins. Vappingo’s free 300-word sample lets you see the approach to your actual writing before deciding on a full order.
Frequently Asked Questions
Does every university require an AI use declaration?
No universal requirement applies to every university and assessment. Some universities prescribe formal declarations, others use different acknowledgements or conditions. Follow the rules for your dissertation.
Do I need to declare Grammarly?
It depends on your institution and the specific Grammarly features used. Ordinary spelling suggestions and generative rewriting are not necessarily treated alike. Follow the relevant definitions and disclosure rules.
What should I write if I used ChatGPT for brainstorming?
If the use was permitted and declarable, name ChatGPT, identify the planning task, explain how you evaluated the suggestions and state whether any generated wording entered the final submission. The examples above provide a starting point.
Should I say no AI-generated text was used if AI only fixed my grammar?
Use more precise wording. State which feature suggested corrections and whether you accepted them. Some features generate new wording, so an absolute claim may be inaccurate even if the intellectual work is your own.
Can an AI watermark appear after proofreading?
A compatible generative model may introduce a statistical watermark when it produces new wording. Small grammatical corrections may leave too little signal to detect, while substantial rewriting may present a greater possibility. Coverage and detection are not universal.
Do I have to declare professional human proofreading?
Human proofreading is not AI use, but some universities require disclosure of third-party assistance or place limits on what proofreaders may change. Check the specific instructions and disclose it in the format requested.
Does a Certificate of Human Editing prove my dissertation was not written by AI?
No. It can document the human editing service described on the certificate. It cannot establish how the original draft was produced, remove existing AI watermarks or guarantee a particular detector outcome.
What if I used AI in a way that was not permitted?
Do not conceal the assistance or assume a declaration makes it permissible. Keep a truthful record and seek advice from the relevant academic or student-support service before submission.
Should I include a declaration if my university does not ask for one?
Do not assume that an unsolicited declaration is always appropriate. Check the submission instructions; if you want to document permitted AI use voluntarily, ask where and how it should appear.
Make the Declaration as Accurate as the Dissertation
An AI use declaration should not be a generic paragraph copied at the last minute. Its purpose is to give an accurate account of what the tools did, what you did and where the academic work came from.
That matters more as AI systems gain the ability to leave invisible provenance signals in supported text. Watermarking does not decide whether you complied with your university’s rules, but it reinforces the importance of distinguishing feedback, grammar correction and generated prose.
And when the dissertation is already yours, the final language check does not have to involve generative rewriting. Vappingo’s human-only proofreading offers visible edits, editorial judgement and a clear record of the work done—provided that the service and level of intervention are permitted by your institution.
Continue reading: University AI Policies Explained · AI vs. Human Proofreading for Dissertations · Can AI-Written Text Be Watermarked?.