AI Detection at Canadian Universities: What Schools Put in Writing

It depends on the school, and at a lot of Canadian schools the answer is written down on a public page you can read in about two minutes. Of the six institutions we read on 27 August 2026, every one says some version of no: not enabled, not approved, not supported, not endorsed, discontinued. Privacy is named in the reasons at four of them. Between the other two the reasons are reliability, bias and cost.

That is different from asking whether Turnitin can detect AI. It can, when the institution licenses it and switches it on. The question is whether your institution did. The schools in this article describe different decision paths: UBC names a Privacy Impact Assessment, Laurier names its own Privacy and Security Impact Assessment process, and Concordia says online AI detectors have not been approved or acquired.

The general mechanics of the detector are covered elsewhere on this site. Whether colleges use AI detectors at all and why some schools show no AI score do that job. This page is about the Canadian layer sitting on top of it.

In British Columbia the rule used to be blunt: student personal information stayed on servers in Canada. That changed. UBC's Faculty of Arts instructional support site describes the current position:

Stemming from the amendments made to FIPPA from Bill 22 - 2021, passed on November 25, 2021, student information can now be held on servers outside of Canada, however, the learning technology storing that information must first pass a Privacy Impact Assessment (PIA). All tools supported by LT Hub have passed a PIA, including Turnitin.

Read the shape of that. The data-residency wall came down and a review gate went up in its place. A tool is not allowed because it is popular or because a professor wants it. It is allowed because it went through an assessment.

Then read what UBC's Learning Technology Hub says about AI detectors specifically, in the post that affirmed its decision:

The use of other AI-detection tools is also not recommended, due to similar issues as those noted above, as well as privacy and security concerns; no AI-detection tool has undergone a UBC Privacy Impact Assessment yet.

The gate and the outcome, in two sentences on two pages of the same university.

At Wilfrid Laurier in Ontario, the institution uses a different named process. Its teaching and learning guidance points at Senate Policy 9.5 and says approval through a Privacy and Security Impact Assessment is required before using any software tool with data that can identify a student. Then it says what that means for an instructor holding your essay:

Please note that assignments and work authored by a student represent the student's intellectual property. Under no circumstances should instructors submit it to any AI tool or third-party software system unless it is approved via the University PSIA process.

Concordia, in Quebec, states the outcome of that process rather than the process:

At the moment, Concordia has not approved of or acquired the use of any online AI detectors and therefore the use of them by staff or faculty is not permitted due to privacy laws and regulations.

These are three institution-level statements, not one Canadian rule. UBC and Laurier explicitly name PIA or PSIA review. Concordia states the outcome it applies: no approved or acquired online AI detector, and staff or faculty use is not permitted because of privacy laws and regulations.

All of these were read on 27 August 2026. Wording changes, so the date matters more than the summary.

Two of those entries deserve their own paragraph.

Waterloo published its reasoning rather than a decision line, and the reasoning includes something a vendor page will never tell you:

Furthermore, internal testing by the University's Instructional Technologies and Media Services (ITMS) group found the advantages of the detection tool were inconclusive. For example, in more than one instance the product flagged human written text as 100% generated by AI, which is concerning.

That is an institution saying, in public, that it watched its own copy of the tool return 100% on human writing. If you have ever been on the wrong end of that number, why a submission comes back as 100% AI explains the mechanism, and Waterloo's page is a useful thing to know exists.

Laurier's sentence goes further than most, because it is not about whether the tool runs. It is about whether the output counts. Detector output `will not be accepted as evidence of academic misconduct`, and the same page says instructors cannot skip the misconduct procedure and hand out their own penalty for suspected AI use. Those are two separate protections and students routinely only hear about the first.

UBC's decision post is worth one more paragraph, because it names a harm that lands on the student rather than on the institution:

Results from the feature are not available for students to review: The report provided by the AI-detection feature in Turnitin is only accessible to instructors; students will not be able view the results. ... That lack of functionality makes it more difficult for students to review information about suspected misconduct and to respond to any errors in flagged content.

The typo and the missing article are UBC's. This describes the default student view, not proof that the feature was disabled: Turnitin's own FAQ says the indicator and report are not visible to students, while an instructor can download the AI report as a PDF and share it. UBC's page also notes that an instructor has nothing to compare a flagged passage against, because for AI detection `the source material simply does not exist`.

A lot of Canadian confusion comes from collapsing two different things into "does my school use Turnitin".

Similarity checking and AI writing detection are separate features with separate licensing and a separate on/off control. Turnitin's own FAQ describes the control:

Yes, admins have the option to enable/disable the AI writing feature from their admin settings page. Disabling the feature will remove the AI writing indicator & report from the Similarity report and it won't be visible to instructors and admins until they enable it again.

That admin switch explains what instructors and admins see, but it does not explain a student's view by itself. Turnitin's FAQ says the AI indicator and report are not visible to students even when the feature is running; an instructor can download and share the PDF. A student who sees no percentage therefore cannot infer whether the feature was licensed or enabled and should ask the instructor or administrator. The separate status `AI writing detection was not enabled at time of submission` establishes disablement only when that status actually appears on a report.

The similarity half often comes with a right the AI half does not. At the University of Toronto the conditions of use for the plagiarism detection tool say instructor use is voluntary, and then:

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 University's plagiarism detection tool, instructors will need to find alternative arrangements to check their work as rigorously.

The same passage adds the sentence people miss: `Students cannot be penalized for choosing to opt out.` The page does not prescribe what the alternative arrangement will be; it only says the instructor must arrange another way to check the work as rigorously.

Waterloo did not just record that it had turned the detector off. Its page gives a reason, and one line of it lands differently on anyone writing in a second language: `AI detection tools have also been found to be biased toward students whose first language is not English`, followed by a citation to the literature.

What makes that sentence useful is not that it proves anything about your own paper. It is where it comes from. It is not a forum post or a blogger's summary. It is a university writing down why it stopped using a tool it had already paid for, on its own site, under its own name.

The evidence behind that claim, including the limits of it, is the subject of AI detector bias against non-native English writers.

Four things, in the order they are worth doing.

Find your own school's page before you argue about the number. The pages quoted above live in predictable places: the teaching and learning centre, the associate vice-president academic, the faculty IT support site, the office of the vice-provost. Search your school's domain for the tool name rather than for "AI detection policy", because the sentence you want is usually inside a Turnitin instructor guide rather than in anything called a policy.

Check the date on it. Waterloo's decision took effect in September 2025. Toronto's AI-detection answer carries a June 2025 update stamp. A page written in 2023 tells you what somebody decided in 2023.

Separate the two questions when you write to anyone. "Does my program have AI writing detection enabled" is answerable. "Does my school use Turnitin" gets you a yes that means nothing.

Know what the privacy rule protects and what it does not. At UBC's Arts faculty the guidance to instructors is that when Turnitin is used as a standalone tool, students must be allowed to create accounts with anonymized emails if they want to, and the page recommends stripping names, student numbers and document metadata before upload. Those are handles you can ask about. What the rule does not do is decide whether AI writing was used in your assignment. It decides which tools are allowed to hold your work.

This is the situation where the Canadian paperwork matters most, and it is worth being precise about what it does for you.

At Concordia, running an online AI detector is `not permitted` for staff and faculty. At Laurier, submitting student work to a third-party system without PSIA approval is ruled out `under no circumstances`, and detector output is not accepted as evidence. At Toronto, the university `does not support` the use of these programs and names sharing student work with them without permission as a privacy and ethics concern.

None of that is a finding about your assignment. It is about the standing of the evidence and the process that produced it, which is a different argument and often a stronger one. If you are heading into that conversation, preparing for an academic integrity meeting and how to challenge a false AI flag cover the part that is about you rather than about policy. The bias question that Waterloo cites in its rationale has its own evidence, which we walk through in AI detectors and non-native English writers.

Ask, in writing, which tool produced the number and whether it is an approved institutional tool. At Concordia, Laurier and Toronto that question has a documented answer before anyone gets to your prose, and UBC's post adds that it does not recommend other AI-detection tools either, partly because none has been through a UBC Privacy Impact Assessment.

Sometimes the outcome is not a dispute at all. An instructor wants a section rewritten, or you want to look at flagged passages before you submit rather than after, which checking your work before submitting covers.

The thing worth controlling there is scope. Every paragraph you touch is a paragraph you have to re-read against your sources, and touching the whole document creates a checking job the size of the whole document. HumanPen can use a Turnitin or iThenticate AI Writing Report as the selection map. Before processing, it displays the complete paragraphs in that map; if a highlight covers only part of one, the boundary expands to the whole paragraph. You confirm that list, unselected text is preserved verbatim, and credits are charged only for the words on it. Where a result qualifies, you can keep lowering AI for free.

HumanPen makes no fixed-score commitment for a later report. What it does control is which words change and which do not.

My Canadian school shows a similarity score but never an AI percentage. Does that mean I am fine? No conclusion follows from the student view alone. Turnitin says the AI indicator and report are not visible to students even when the feature is running, though an instructor can share the downloaded PDF. Ask the instructor or administrator whether the feature was enabled; only an explicit report status saying it was not enabled establishes that state.

Is Turnitin allowed in Canada at all, given the servers are in the US? UBC's Arts IT page says that after the 2021 FIPPA amendments student information can be held on servers outside Canada as long as the technology has passed a Privacy Impact Assessment, and that Turnitin has passed one there. SFU's page says it updated its Turnitin protocols effective January 2012 because data is stored and accessed on US servers, so its practices continue to comply with BC's privacy legislation. Both are institution-level statements about their own deployment, not a national rule.

Can I refuse to have my paper submitted? At the University of Toronto, yes, for the plagiarism detection tool, and the conditions of use say you cannot be penalized for it, though the instructor will arrange another way to check the work as rigorously. Other schools handle this differently. Ask early, ideally in the first week, not the night an assignment is due.

Does opting out apply to AI detection too? The opt-out language I read is written about the plagiarism detection tool. Where a school has not enabled AI writing detection, there is nothing to opt out of. Where it has, ask the department directly rather than assuming the same clause covers both, because the AI Writing Report and the Similarity Report are separate reports.

My professor pasted my essay into a free AI detector. Is that allowed? It depends on the school, and at some of them it is answered in writing. Concordia says use of online AI detectors by staff or faculty is not permitted. Laurier says instructors should not submit student work to any third-party system without PSIA approval, and that output from detection tools will not be accepted as evidence of misconduct. Look for your own school's version before you build an argument on it.

Does a school dropping the tool mean AI use is now allowed? No. Waterloo's page keeps the misconduct framing while dropping the detector, and Laurier's page calls unapproved or unacknowledged generative AI use academic misconduct in the same breath as refusing detector output as evidence. The rule about what you may use did not move. The evidence standard did.