Can My University Use an AI Detector to Accuse Me of Cheating?

The short answer. At the University of Victoria, no, and the no is written into policy. UVic's September 2026 Academic Integrity Policy says instructors may not use generative AI tools to grade work or to decide whether a student committed a violation, and bars the tools from investigations entirely. Other institutions have reached similar positions by different routes.

If you are a student, the question in the title is one you probably already have, even if you have not had reason to ask it out loud yet. A professor runs your essay through a detector. The detector returns a high score. What happens next, and does the score mean what the professor thinks it means?

The answer depends on where you are studying. At one Canadian university, the question is now answered in writing, and the answer is that a detector result cannot be used at all.

What does UVic's policy actually say?

The University of Victoria's new Policy on Academic Integrity came into effect in September 2026. Among the questions its FAQ page addresses directly is whether instructors may use AI-detection tools in misconduct cases.

The policy's answer, quoted from the university's own page: "UVic does not permit instructors to use Generative Artificial Intelligence Tools (GAITs) to grade students' work or to determine whether a student has violated the Policy on Academic Integrity. These tools are known to produce unreliable and often inaccurate results, and [their use] violates the university's privacy policy. As such, they are not permitted as part of any investigation or as evidence when considering a potential violation."

That is two separate prohibitions in one paragraph. An instructor cannot use an AI tool to grade your work. An instructor cannot use one to decide you broke the rules. And even if they wanted to use one as supporting material, they cannot: the tools are not permitted as part of any investigation or as evidence.

The FAQ then addresses the future: "If the university ever chose to allow AI-detection software in the future, this would represent a significant change to how academic integrity is assessed. For that reason, such a decision would need to go to Senate for approval, ensuring that any shift in practice is transparent, deliberate and aligned with the policy." Reversing this position is not a matter for an individual department or course convenor. It requires a senate vote.

Why does UVic give two reasons, not one?

Most critiques of AI detectors stop at accuracy: one study reports a tool is right most of the time, another reports it is wrong more often, and the argument stays on the number.

UVic's policy offers a second and separate reason: privacy. A detection tool receives student writing. Where that text goes, how it is stored, who can see it, and whether it contributes to future model training are questions whose answers differ by vendor and are not always disclosed in plain language. Our piece on what happens to text you paste into an AI detector goes through what the major tools say on this point. For a university with obligations under Canadian privacy legislation, the analysis differs from a solo teacher trying a free tool, but the question is the same.

The policy also captures a dynamic that overall accuracy numbers miss: even a tool that is right most of the time can be wrong in ways that fall disproportionately on particular groups of students. A 2023 study by Liang and colleagues tested seven detectors on essays by non-native English speakers and found that an average of 61.3 per cent of non-native essays were flagged as AI-generated, while native-speaker essays were flagged at near-zero rates. Mount Royal University in Calgary cited the same study in its August 2026 advisory recommending that the university avoid using AI detectors in misconduct processes at all.

Is UVic alone in this position?

Not quite, though a written prohibition is unusual. Several institutions have reached a similar end point through different reasoning.

The University of Manchester's current AI guidelines state that detector output "cannot currently be used as evidence of malpractice." That is a standard-of-evidence argument: the tools are too unreliable to meet what the institution requires. It is not an explicit prohibition on using the tools in the first place.

A New York court reached a similar limit from outside academia. In January 2026, in Matter of Newby v Adelphi University, a state court annulled an academic integrity finding that rested on a Turnitin score of 100 per cent. The court did not rule that the detector was wrong. It found the university had failed to consider the student's written response, had denied him access to an advisor its own rules promised, and had let the same official make the finding and then hear the appeal. The full case is covered in what a court needed beyond a 100 per cent AI score.

The European Network for Academic Integrity's Dubai Accord, launched on 27 August 2026, says "no automated detection output or probabilistic indicator should be treated as conclusive evidence of misconduct", and calls on institutions to give students notice of concerns, access to the substance of the evidence against them, and a route to independent review or appeal.

What UVic adds is a written prohibition rather than a standard-of-evidence limit. The policy does not say a detector result is insufficient on its own; it says the result may not be used at all. That is a stronger statement, and it is now in the document every UVic student and instructor can read before term starts.

What does "not permitted as evidence" mean in practice?

For UVic students, it means an instructor who runs your essay through a detection tool and gets a high score has a signal they cannot formally act on. They cannot open a misconduct investigation on the basis of that score, and they cannot present it to an academic integrity panel as grounds for a finding against you.

That does not mean an instructor cannot notice that an essay seems unusual, or ask you about your writing process, or pay closer attention to how a piece sits alongside your other work in the course. A detector result you never saw might still shape how carefully your work is read. But the formal process, the finding, the sanction, the record on your file, those cannot rest on a tool the policy prohibits.

What should I do if I am told a detector flagged my work?

The UVic policy is useful if you are at UVic, but it does not automatically protect you if your institution has no equivalent rule. Wherever you study, a few practical steps matter.

Ask what the finding actually is. There is a difference between an instructor who noticed something unusual and wants to discuss it, and a formal allegation under your institution's academic integrity procedure. The first is a conversation; the second is a process with defined rules about who decides, what evidence is weighed, and how you can respond. Ask which one you are in before you reply in writing.

Gather your process evidence. Draft history in Google Docs, planning notes, an early outline, messages to a tutor, a version you sent to a friend for feedback. These are things a detector cannot produce and a conversation can. In the Adelphi case, the essay had been written with a tutor, and the court found the university had not considered the student's written response. Assembling that evidence before a meeting is better than hunting for it after.

Read your institution's own rules. Most students do not look at their academic integrity policy until they need it. The document that governs your case is your university's, not a national standard, and it will say what evidence is required, what process you are entitled to, and whether you can bring a support person to meetings. Several universities now require instructors to state permitted and prohibited AI uses in course outlines before the term starts. Our piece on what AI use is allowed in your assignment covers how several institutions have put this in writing.

Understand what a detector result is. No detector, including the one on this site, produces proof. Every result is an AI-pattern score: a signal that the submitted text shares features with AI-generated writing, at a particular threshold and with a particular error rate. Our current model wrongly flags 3 of 400 real student essays at the flag threshold, a figure published with our full methodology in our study on how often AI detectors flag human writing. A high AI-pattern score means your text shares features with AI output. It does not say why, and it does not say what you did.

Where that leaves a detection vendor

We sell an AI detector, so the straightforward reading of a policy like UVic's is that it limits how our tool can be used there, and we think that limit is right.

Our published figures show both what detectors can do and where they fall short. At our current flag threshold we catch about 98 in 100 essays submitted as direct chatbot output, 69 of 70 lightly reworded ones, and about 81 in 100 with typos and contractions added. We catch 26 of 70 essays that were half AI and half the student's own writing, the case that matters most when a student genuinely mixed their own thinking with an AI draft, and about 19 in 100 when the chatbot was asked to write like a 15-year-old, which is where we are weakest. All figures, including the cases we miss, are on our methodology page.

A screening result can tell a teacher which essay to read more carefully, and which student to speak to. UVic's policy makes explicit something that belongs in the framing of every result regardless of institution: the conversation that follows is the process. The AI-pattern score is the starting point for it.

Who wrote this, and what we sell. Is It AI is an AI-writing detector for teachers and students. All policy quotations above are from UVic's own pages, linked below. We have no connection to the University of Victoria.

Sources

• University of Victoria, Policy on Academic Integrity FAQ, in force September 2026. Policy text quoted directly from this page.

• University of Victoria, Supports for UVic's New Academic Integrity Policy, VP Academic, September 2026.

• Supreme Court of New York, Nassau County, Matter of Newby v Adelphi University, 2026 NY Slip Op 26021, decided 28 January 2026.

• CHEK News, Integrity in the age of AI: UVic updates academic misconduct policy.

• Times Colonist, UVic can now punish AI cheaters, even after graduation.

• Liang et al., GPT detectors are biased against non-native English writers, 2023, cited via Mount Royal University advisory.

• University of Manchester, University AI Guidelines.

• European Network for Academic Integrity, Dubai Accord on Academic Integrity in the Age of Artificial Intelligence, 27 August 2026.