At McGill, it depends on the use, and you check with your supervisor and department before you start. Graduate and Postdoctoral Studies (GPS) sorts uses of generative AI from low risk to very high risk. Since May 1, 2026, it has recommended a GenAI statement and a disclosure framework for master's and doctoral theses. GPS says these may become mandatory, so check the current version.
GPS asks you to consult your supervisor, your department, and its guidelines on whether, how, and when you can use generative AI, and to do that before you start the research and the thesis. If a journal's rules differ from McGill's, GPS says McGill's guidelines apply to your thesis.
The statement itself is a recommendation for now: “The use of a generative AI statement and disclosure framework is recommended, but not yet mandatory.” GPS also gives a reason to use it anyway: “Failure to disclose the use of generative AI tools in a submitted thesis may compromise and/or delay assessment of the thesis, as examiners have been actively identifying and flagging potential generative AI use.”
GPS says so in its FAQ: “Following consultations with committees and departments in Fall 2026, the statement and disclosure frameworks may become mandatory”. Those consultations are scheduled for fall 2026, so look at the GPS thesis page before you submit. The version this page describes is dated May 1, 2026, and we read it on November 4, 2026.
GPS offers two fixed versions of the statement, placed before the main body of the thesis, and asks you not to modify them. If you want different wording, you can raise it with your department, which can recommend changes to GPS. The disclosure framework is flexible, and you choose it with your supervisor.
| Document | What it covers |
|---|---|
| Statement, if you used GenAI | You consulted the GPS guidelines and your supervisor, and you reference the use where necessary |
| Statement, if you didn't | You consulted the GPS guidelines and didn't use generative AI in producing the thesis |
| Framework: text-based declaration | Tools used, purpose, extent, human oversight and verification, and impact on the research |
| Framework: checklist | Boxes for language editing and references, coding help, identifying articles, or other uses |
| Framework: checklist with extent | The same uses, each rated minimal, moderate, or high |
The November 2024 guidelines divide uses into three categories. They address research, and GPS says they aren't course-based. The examples in the table come from the GPS lists.
| Category | One example from the guidelines | What GPS says about it |
|---|---|---|
| Low to moderate risk | “Search engines: Using generative AI to sift through and identify relevant research articles for a project” | It supports work you are already doing. Verify every reference, since some may not exist |
| Moderate to high risk | “Literature review: Producing summaries of published research on a given topic for one’s own consumption” | The content is for your own use, and GPS calls for extra care in this category |
| Very high risk | “Research writing: Writing entire sections or the entire proposal/paper” | Presenting generated content as your own work counts as plagiarism under McGill's Code of Student Conduct |
GPS says work that is only partly derived from generative AI could also lead to a violation. It encourages supervisors to set clear expectations on how these tools are used in research, and it says the permissibility of some uses, such as translation, depends on your field. Treat the categories as a starting point for that conversation with your supervisor.
This example is made up. A doctoral student in epidemiology asks a tool to summarize forty articles, to decide which ones to read. On the GPS lists, that falls under moderate to high risk, as a use for the student's own reading. Later, one summary, lightly reworded, goes into the literature review chapter. Submitting such a summary as your own work is on the very high risk list. To fix this, the student goes back to the article, reads it, drafts the passage from that reading, and mentions the earlier use in the framework.
Esmond evaluates every assignment before you submit. It never writes it.
Esmond is an AI tool that doesn't write for you. If your course doesn't allow any AI tools, don't use Esmond for that work. If you're unsure, ask your instructor or advisor.
If you use Esmond for comments on a draft, treat it like any other tool on your framework: bring it up with your supervisor, and list it if your framework lists tools. McGill has not reviewed or approved Esmond.
To see what you'd list on your framework, read how Esmond works and open the sample report. Also see whether you can use Esmond.
GPS asks you not to modify the main statement. You can send suggestions through your department. The disclosure framework is flexible.
GPS gives a separate version for that case, which you may include before the main body of the thesis.
The guidelines place spell and grammar checks in the low-to-moderate category. Your supervisor or department may still have a view, so ask if you are unsure.
GPS calls them research-focused and not course-based. In a course, follow your instructor's rule.
Read on November 4, 2026.
Esmond is not taking orders on this site yet.