Plagiarism and AI at U of T: your right to opt out of the detection tool

The short answer

At the University of Toronto you can decline to submit your work to the plagiarism-detection tool, and you cannot be penalised for it. That right is published, unconditional, and backed by a rule that stops instructors from submitting on your behalf.

It is the most substantial student protection we have found at any university in this series, and almost nobody writes it up — so most U of T students do not know it exists.

The opt-out right, in full

Start with the syllabus statement. U of T requires instructors who use the tool to include specific wording, and it “cannot be altered in any way”:

“Normally, students will be required to submit their course essays to the University’s plagiarism detection tool for a review of textual similarity and detection of possible plagiarism. In doing so, students will allow their essays to be included as source documents in the tool’s reference database, where they will be used solely for the purpose of detecting plagiarism.”

That first word carries the whole thing. Normally — not always. And here is what it opens:

“Students are permitted, under our conditions of use, to opt-out of using the University’s plagiarism detection tool. If a student chooses not to submit their assignment through the plagiarism detection tool, instructors will need to find alternative arrangements to check their work as rigorously. (It should be noted that very few students choose to opt out.) Students cannot be penalized for choosing to opt out. If students choose to opt out, they should let their instructor know well in advance of submitting their paper. Ideally, they should communicate this during the first class, when the instructor is reviewing the course outline.”

The mechanism that makes it real is the submission route. Students submit for themselves, which means nobody can quietly do it for you:

“Instructors cannot submit a paper through University’s plagiarism detection tool without the student’s knowledge. You can ask your student to submit directly to the University’s plagiarism detection tool but you cannot require them to do so.”

What happens instead? U of T publishes the alternatives: “requiring annotated bibliographies from students; requiring students submit all rough work with their papers; requiring that students include the call numbers or web site addresses of all sources cited in their paper.” Or a short meeting with “pointed questions about their research methodology”. And there is a limit on how much extra this can cost you: “Instructors should not be asking students to do anything that might be perceived as additional work.”

Two practical notes. Say so early — the guidance says first class, not the night before the deadline. And be realistic about the trade: you are choosing a different form of scrutiny, not less of it. Keeping your rough work and version history from the start makes that route painless, which is the same habit we recommend in how to prove you wrote it yourself.

On ownership, U of T is also clearer than most: “Under Canadian copyright laws students retain the copyright to any papers submitted”, submitted papers “are not the property of” the tool, instructors “can only access those papers that were directly submitted to their own course”, and the agreement “ensures that student papers […] will not be used for commercial purposes.”

One gap worth knowing: where the data is stored is not published. U of T says papers are held “in the company’s database” but no page we found states the server jurisdiction, and no Ontario freedom-of-information analysis of the tool is published. If that is your reason for opting out, you are entitled to ask.

Graduate theses need explicit consent

A separate and stronger rule applies at the top of the degree ladder, and it is one line:

“Supervisors should ensure that they have the explicit consent of a student before they submit their thesis to University’s plagiarism detection tool.”

For coursework, the default is submission with an opt-out. For a thesis, the default is the other way round: consent first. If you are writing a thesis at U of T and it has been submitted without you being asked, that is not how the conditions of use read.

The tool U of T never names

This is a small oddity worth reporting accurately. Across its own pages, U of T calls it “the University’s plagiarism detection tool” — never the product name. The only identifying trace is a line about the licence: “The intellectual property of all students submitting to University’s plagiarism detection tool is protected by the licensing agreement between the University of Toronto and iParadigms.” We will report that and no more; if you want the vendor confirmed, ask your division.

Three facts about how it is used:

• Instructor use is optional: “The use of the University’s plagiarism detection tool by our instructors is completely voluntary.” Many courses never touch it.
• You cannot self-check.Submission is through a Quercus assignment your instructor created, and “instructors and students should not access the program through the company’s website.” You can see only your own submissions.
• The report is not a verdict: “Similarity Reports highlight passages that may be of concern, but do not automatically report whether a paper has been plagiarized. Determining if plagiarism has occurred requires analysis by the instructor.” On what that percentage does and does not mean, see what a similarity score actually means.

U of T also describes the integration as providing “textual similarity analysis only” — which is the closest thing it publishes to a statement about configuration.

How AI enters the Code

Here is something that surprises people: the Code of Behaviour on Academic Matters — approved by Governing Council on 24 June 2025 and effective 1 July 2025 — does not contain the words “artificial intelligence”, “generative AI” or “ChatGPT” anywhere. AI is not a named offence at Toronto.

It arrives through an older category instead. The Provost’s office:

“If an instructor specified that no outside assistance was permitted on an assignment, the University would typically consider a student’s use of generative AI to be use of an ‘unauthorized aid’ under the Code of Behaviour on Academic Matters.”

And the baseline, which is stricter than many students assume:

“The University expects students to complete assignments on their own, without any outside assistance, unless otherwise specified.”

So permission comes from your course, not from the absence of a prohibition. U of T acknowledges the confusion this creates and puts the duty on staff: “students might receive a range of instructions across different courses about what constitutes appropriate use of AI within each course. We therefore encourage all instructors to be very transparent and clear.”

The learning-aid carve-out

This passage is more permissive than almost anything else in this series, and it is worth reading carefully because it has a boundary:

“Undergraduate and graduate students may use GenAI tools as learning aids, for example, to summarize information or test their understanding of a topic. These tools should be used in a manner similar to consulting library books, online sources, peers, or a tutor. Such uses are generally acceptable even if an instructor has stated that AI tools are not otherwise permitted in the course. These uses typically do not need to be cited or disclosed.”

Using a chatbot to understand a concept before you go and read the real sources is treated like asking a tutor. That is a sane position, and rarer than it should be.

Then the line:

“However, if students use information from GenAI tools that would normally require citation (e.g., quoting, paraphrasing, or reproducing text or ideas), they must cite their use according to guidelines provided by U of T Libraries or as directed by their instructor.”

Understanding is free; using the output is a citation. And instructors may go further: “Instructors may require students to describe their use of GenAI in assignments, but this must be clearly stated in the syllabus and assignment instructions.”

There is no single university-wide disclosure wording at U of T — the Libraries publish a citation guide covering MLA, APA and Chicago, and the form is course-dependent. Our AI disclosure statement guide covers what a good one contains. Whatever you cite, check it exists first: fabricated references are the most common way an honest student ends up in a process, and verifying AI citations walks through it.

What the Provost’s office tells instructors

If you are ever accused on the strength of a detector output at Toronto, this is the passage to bring. It is from the Office of the Vice-Provost, Teaching & Learning, updated 18 June 2025:

“The University does not support the use of AI-detection software programs on student work. None of these software programs have been found to be sufficiently reliable, and they are known to incorrectly flag instances of AI use in human-written content. […] In addition, sharing your students’ work (i.e., their intellectual property) with these software programs without their permission also raises a range of privacy and ethical concerns.”

Note that there are two independent grounds — reliability and privacy — and the second one stands even if the first were someday solved.

The teaching centre adds a point that matters for anyone writing in a second language: AI-detection programs are “unreliable and biased against non-native English writers”, and it notes that human intuition fares no better — “Personal intuition that a text is AI-generated has been shown to be inconsistent, even when evaluators are experienced with GenAI and confident in their abilities.” We wrote about that asymmetry in AI detectors and non-native speakers.

What U of T recommends instead is human: “meeting with a student to discuss their work in person, using short in-person assessments that ask students to expand on the ideas in their out-of-class assignments”.

One honest limit: “does not support” is a statement of policy, not a confirmation that any particular indicator is switched off in the software. Whether the AI-writing indicator is technically disabled in the Quercus integration is not publicly documented.

And to be consistent about our own tool: the same caution applies to us. Our published error rates are measured on English and German corpora, and no detector produces proof. See how accurate AI detectors really are.

Clear and convincing evidence

Most universities decide these cases on the balance of probabilities — more likely than not. Toronto sets the bar higher, and this is written into the Code:

“The onus of proof shall be on the prosecutor, who must show on clear and convincing evidence that the accused has committed the alleged offence.”

Two things follow. The burden is on the university, not on you. And “clear and convincing” is a materially stronger standard than “more likely than not” — which is difficult to satisfy with a probabilistic score from a tool the university has said it does not support.

If you are facing this, what to do when you are accused of using AI sets out how to assemble the material that ends it fastest.

If a detector score is what started this, our guide to an AI detector flagging your thesis sets out what that number does and does not establish on its own.

Sanctions

At divisional (dean’s) level: an oral or written reprimand; a reprimand with resubmission for minor first offences; zero or failure for the piece of work; a grade reduction in the course; denial of facility privileges; a monetary fine; “assignment of a grade of zero or a failure for the course”; and suspension “for a period of not more than twelve months”.

At University Tribunal level: any of the above, plus a zero or failure “for any completed or uncompleted course or courses in respect of which any offence was committed”, suspension of up to five years, a “recommendation of expulsion from the University”, and a recommendation to Governing Council “for cancellation, recall or suspension of one or more degrees, diplomas or certificates obtained by any graduate”.

Sanctions can be recorded on the academic record. We are not going to quote a duration here, because we were not able to verify the exact wording of that clause against the Code itself — ask your division, or read the Code directly, rather than trusting a number from anyone who has not.

The Tribunal hears a matter where the student denies the offence, where the sanction sought exceeds the dean’s authority, where there is a prior offence, or where multiple offences are alleged. Appeals go to the Discipline Appeals Board within 21 days, on questions of law or as to sanction.

U of T reports annually to Governing Council on the operation of the Code, but we did not retrieve that report, so we are not giving you case counts. Anyone quoting one should be able to point at the document.

Before you submit

• Decide about the opt-out in week one, not at the deadline. That is what the guidance asks for.
• Keep rough work. It is the published alternative to the tool, and it is also your best evidence if anything is ever questioned.
• Assume no outside assistance unless your course says otherwise.That is Toronto’s stated baseline.
• Understanding a topic with a chatbot is fine; using its output is a citation. That is the line the Provost draws.
• If you are writing a thesis, your consent is required before it goes through the tool.

If you want a read on your own text before you hand it in, our AI check is free up to 1,500 words, with no account and no name — and nothing you submit is added to a third-party comparison database. Our published error rates are measured on English and German corpora, and no detector, ours included, produces proof.

Sources

• University of Toronto: Code of Behaviour on Academic Matters — approved by Governing Council 24 June 2025, effective 1 July 2025.
• Centre for Teaching Support & Innovation: Plagiarism detection — conditions of use — last modified 19 April 2025; the opt-out and the graduate-thesis consent rule.
• Office of the Vice-Provost, Teaching & Learning: Generative Artificial Intelligence in the Classroom: FAQ — updated 18 June 2025.
• Centre for Teaching Support & Innovation: Teaching and Generative AI at U of T — teaching tips.

This page is orientation, not legal advice. What binds you is your course syllabus and the Code of Behaviour on Academic Matters.