Short answer: yes, at most Canadian universities you may use generative AI in your thesis — but only with the prior approval of your supervisor and supervisory committee, only if you document and describe that use inside the thesis itself, and only if you can defend every part of it orally. Unapproved use is an academic offence even when you cite the tool.
The question is usually asked as though Canada had an answer. It does not. There is no federal rule, no provincial rule, and no national accreditation body that fixes what a graduate student may do with ChatGPT or any other generative tool. Degree requirements in Canada are set institution by institution, so the rule that binds you is the one written by your own school of graduate studies — and then narrowed further by your supervisory committee.
What makes the question answerable anyway is that the largest Canadian graduate schools have now published guidance, and it converges. Read two of them side by side and the same three-part test appears in both.
What does the University of Toronto actually require?
The School of Graduate Studies publishes Guidance on the Appropriate Use of Generative Artificial Intelligence for Graduate Academic Milestones, last updated 17 July 2026. It covers comprehensive exams, candidacy, language requirements, the doctoral thesis and the final oral examination.
Four sentences in it do most of the work:
- Prior approval, not after-the-fact disclosure. “The use of generative AI tools in any aspect of researching or writing of the thesis must be done with the prior approval of the supervisor(s) and supervisory committee.”
- The description goes in the thesis. “Careful attention must be paid in the thesis to appropriately citing and describing any use of generative AI tools in the research process.”
- Unapproved use is an offence. “Unauthorized use of generative AI tools for scholarly work at the University of Toronto may be considered an offence under the Code of Behaviour on Academic Matters.”
- You defend it out loud. “Students must be able to describe and defend any use of generative AI, as well as the contents of the thesis during their final oral examination.”
The guidance also sets the test the committee has to be able to apply: “it must be clear how the student’s versus the AI tool’s contributions will be identified, and it must be possible for the student to provide sufficient evidence that they themselves have met the Key Criteria.” That is the operative sentence for anyone deciding how to work day to day, because it is a requirement about evidence, not about intention.
What does UBC require, and how is it different?
The University of British Columbia’s Graduate School reaches the same destination by a slightly different route. With the approval of their supervisor and committee, graduate students may use generative AI — and students must “proactively obtain such approvals, and document in writing any use of GenAI” in their work.
UBC then does something U of T does not: it names the physical location of the disclosure. If generative AI was used in the research, the drafting of, or the preparation of a thesis or dissertation, the Preface must include “a concise description of how it was used.” And if it was not used at all, the Preface must carry “a clear statement that generative AI was not used for any aspects of the work.”
Read that second half again, because it catches people out. At UBC the disclosure is not optional in the sense of “say something if you used it.” A thesis that used no generative AI still has to say so. Silence is not a valid state.
UBC is also explicit about the consequence, and the phrasing matters: use of generative AI “not approved by the student’s supervisory committee or graduate program (even if appropriately cited) may legitimately result in changes being required to the work, the student’s work not being approved for defense, or other actions” under the academic misconduct policies.
“Even if appropriately cited.” Citing the tool does not retroactively create permission. Students routinely assume that a transparent acknowledgement in the front matter converts unauthorised use into authorised use. At UBC, in writing, it does not.
What is the common three-part test?
Strip the institutional vocabulary away and both universities are asking the same three questions, in the same order.
| Test | What it means in practice | When it has to be satisfied |
|---|---|---|
| Prior approval | Your supervisor and committee agreed to the specific use before you did it | Before the work starts |
| Documented, separable contribution | You can show which words, analyses or ideas are yours and which came from a tool | Continuously, while you work |
| Oral defensibility | You can explain and justify every use in the examination room | At the defence |
The middle row is the one that quietly determines whether you pass the other two. If you cannot reconstruct which passages a tool touched, you cannot describe the use accurately in your front matter and you cannot answer a committee member who asks about a specific paragraph. That is a records problem, not an ethics problem, and it is solved months earlier than the defence.
Is any use categorically forbidden?
Nothing in either guidance document bans generative tools outright. What both do is make the intellectual contribution non-delegable. The reasoning, the argument and the interpretation have to be yours; the committee has to be able to verify that; and the degree is awarded for your work.
There is a useful precedent here that predates AI entirely. U of T’s School of Graduate Studies already permits an external copy editor only with prior approval and a strong rationale, requires that the intellectual content of the thesis cannot be modified through the external editing process, and requires the student to identify the editor’s contribution in the thesis document. The generative-AI rule is that rule, re-applied to a new kind of assistance. If you already understood the editing rule, you already understand this one — a point that sits alongside the wider question of what each Canadian credential demands, covered in our guide to Canadian graduate requirements: thesis, dissertation or major research paper.
Does the rule change if I only use AI for editing or translation?
Not automatically, and this is where most of the real risk sits. Both universities frame the requirement around any use in researching or writing the thesis, not around a category of use. Grammar correction, translation, restructuring a paragraph, summarising a source, generating code for an analysis and drafting prose are all uses. Whether each one needs its own conversation is a decision for your committee, which is exactly why the guidance tells you to have the conversation before you start rather than to guess where the line falls.
The practical move is to bring a written list to a supervisory meeting — the specific tools, the specific tasks, the specific chapters — and get the answer recorded in the meeting minutes. A committee that has approved a defined scope in advance is a committee that cannot be surprised at the defence.
How should I write the disclosure statement?
Follow your own institution’s template if one exists. Where none does, the elements both guidance documents point at are these:
- Name the tools. The product and, where you can, the version and the period you used it.
- Name the tasks. “Grammatical correction of Chapters 3 and 4” is a description; “assistance with writing” is not.
- Name what was excluded. State plainly which parts of the work — data collection, analysis, interpretation, argument — involved no generative tool.
- Record the approval. Note that the use was approved in advance by the supervisor and supervisory committee, since that is the fact the reader is checking.
- Keep it verifiable. Everything in the statement should be something you could substantiate if you were asked for drafts.
Where a tool helps most is in the boring, verifiable part of that list: keeping chapter structure, sources and citations consistent so that what is yours stays legible as yours. Draft your chapters in Tesify — the scaffolding and referencing stay consistent while every argument remains 100% written by you.
What about research ethics and AI?
Generative-AI approval and research-ethics approval are separate gates and neither substitutes for the other. If your work involves human participants, the ethics requirement attaches to what you do to those participants, and it has to be cleared before the research begins — see our guide to REB approval and the TCPS 2 rules in Canada. Feeding participant data into a third-party tool is a data-handling question your Research Ethics Board will want to hear about, and it is not answered by your supervisor’s sign-off on writing assistance.
Where do the rules for coursework differ?
Thesis milestones and course assignments are governed by different documents. In a course, the instructor’s syllabus is the binding statement; U of T’s academic integrity guidance notes that if instructors plan to use the plagiarism detection tool for any of their assignments, they must inform their students at the start of the course. In a thesis, there is no syllabus — the supervisory committee performs that role, which is why the approval has to be requested rather than read off a page.
One consequence worth stating: an instructor’s permissive AI policy in a course you took last term tells you nothing about what your committee will accept in Chapter 4.
What if my department has not published anything?
Many have not, and that is not permission. In the absence of a departmental statement, the institution-level guidance applies, and where that too is silent the conservative reading is the one that survives a defence: get approval in writing, keep records of what the tool did, and disclose it. A committee cannot penalise you for having been more transparent than required.
If you are also choosing analysis software for the same project, the licensing and reproducibility questions there are related but separate — our comparison of SPSS, R, jamovi and Stata for a Canadian thesis covers what your institution may already provide.
Frequently asked questions
Can I use ChatGPT for my thesis in Canada?
At most Canadian universities yes, with the prior approval of your supervisor and supervisory committee, provided you document the use, describe it in the thesis, and can defend it at your oral examination. Unapproved use may be treated as academic misconduct.
Is there a national Canadian rule on AI in theses?
No. Graduate degree requirements are set by each institution through its school or faculty of graduate studies. The binding rule is your own university’s, then your committee’s.
Does citing the AI tool make the use acceptable?
Not on its own. UBC states that unapproved use “even if appropriately cited” may still lead to required changes, refusal to approve the work for defence, or misconduct proceedings.
Where does the disclosure go in the thesis?
At UBC, in the Preface. U of T requires the use to be cited and described in the thesis without prescribing a single location. Follow your own school of graduate studies’ formatting guide.
Do I have to say anything if I did not use AI at all?
At UBC, yes — the Preface must contain a clear statement that generative AI was not used for any aspect of the work. Check whether your institution has the same requirement.
Will I be asked about AI use at my defence?
You may be. U of T’s guidance states that students must be able to describe and defend any use of generative AI, as well as the contents of the thesis, during the final oral examination.
Does using AI for grammar checking count?
Both universities frame the requirement around any use in researching or writing the thesis rather than around a category of use. Ask your committee about the specific tasks rather than assuming editing is exempt.
What if my supervisor says yes but the committee was not asked?
Both guidance documents name the supervisory committee alongside the supervisor. Get the approval recorded at a committee meeting so the record exists independently of one person’s memory.
Does AI use affect my research ethics approval?
It can. Ethics approval concerns what you do with participants and their data. Putting participant data into a third-party tool is a data-handling question for your Research Ethics Board, separate from writing assistance.
What happens if I already used a tool without asking?
Raise it with your supervisor before submission rather than after. A disclosure made voluntarily and early is a very different conversation from one that begins in the examination room.
Is departmental silence the same as permission?
No. Where a department has published nothing, the institution-level guidance still applies, and the safe course is prior written approval plus disclosure.
Are the rules the same for a master’s and a doctoral thesis?
The published guidance at both universities addresses graduate work generally, including master’s theses and doctoral theses. Programme-level requirements can be stricter, never looser.