# What AI Detector Do Universities Use? Tools, Policies and Limits · ScriptGrain > Usually Turnitin's AI writing indicator, which comes with the plagiarism software universities already license. Some opted out or switched it off, and a score alone cannot prove misconduct. Canonical: https://scriptgrain.com/blog/what-ai-detector-do-universities-use # What AI Detector Do Universities Use? Tools, Policies and Limits *By Jack Stovell [X](https://x.com/StovBuilds) · [Instagram](https://www.instagram.com/stovbuilds) · [LinkedIn](https://www.linkedin.com/in/jackstovell/) · 2026-10-10 · Guides* Short answer: if a university uses a detector at all, start by assuming Turnitin, then check. It is the obvious choice because it comes with software the university already pays for. Whether the AI feature is switched on, who can see it, and what a score may count as are separate questions, and each university decides them for itself. ## The short answer: Turnitin, usually Three questions get mixed up in most answers to this: 1. **Which tool is licensed.** For most universities with a plagiarism checker, that is Turnitin. 2. **Whether the AI feature is switched on.** Licensing Turnitin does not mean the AI indicator is in use. Some universities opted out at launch, and some have switched it off since. 3. **What a score is allowed to count as.** Even where the indicator is on, policy decides whether a high score triggers anything at all. We can't tell you that every university in a country, or every member of a group like the Russell Group, uses Turnitin's AI indicator. Policies differ and change, so take "Turnitin, usually" as the place to start checking. ## Why Turnitin is the default (it is already in the VLE) Most of it comes down to procurement. Universities already pay for Turnitin's similarity checking, and it already sits inside the virtual learning environment where students hand in work. Adding an AI indicator to that is a small step. Buying, vetting and connecting a separate detector is a much bigger one. Convenience has a cost, though. When the AI indicator sits beside a similarity score in the same report, the two can look like the same kind of evidence. A similarity match points at a source you can open and compare. An AI score points at nothing you can inspect. It is a statistical estimate of how the text was produced. Turnitin publishes its own accuracy claims, including a document-level false positive figure with conditions attached. Read those conditions in Turnitin's current guidance before relying on the number. Our [AI detector accuracy reference](https://scriptgrain.com/reference/ai-detector-accuracy) sets vendor claims against independent testing for Turnitin and other tools. ## Other detectors universities mention: GPTZero, Copyleaks and others Standalone detectors such as GPTZero and Copyleaks are sold separately from Turnitin. Whether a university allows staff to use one is a policy decision. When Ulster University's working group was drawing up guidance in 2023, it noted that some staff had been using web-based detectors before any tool was institutionally approved, which is part of why universities write rules on this at all. We have not listed which universities allow which tools. That would mean checking every institution's published policy, and an unchecked list would be guesswork. If you want to know how these tools reach a score, the [explainer on how AI detectors work](https://scriptgrain.com/reference/how-ai-detectors-work) covers the methods. ## Some universities switch detection off Tool lists tend to leave this out. When Turnitin released the AI feature in April 2023, a number of universities chose not to use it. - **University of Manchester** (5 April 2023) opted out of the first release "along with the majority of UK universities", in its own words. It said Turnitin could not satisfactorily explain how the tool works or how its accuracy figures were calculated, and that launching at the start of the assessment period was "particularly problematic". It presented the opt-out as time to evaluate the tool, with no permanent decision made. - **Bournemouth University** (13 April 2023) also opted out. It pointed to short notice, no chance to test the tool, and a need to fit it into its academic offences policy first. - **Vanderbilt University** in the US (16 August 2023) disabled the detector outright. Taking Turnitin's claimed 1% false positive rate and the 75,000 papers Vanderbilt submitted in 2022, it estimated that around 750 papers could have been wrongly labelled as partly AI-written. It also cited bias against non-native English writers and the lack of any explanation of how the tool works. - **Curtin University** in Australia announced in September 2025 that it would disable Turnitin's AI writing detection from 1 January 2026, keeping ordinary text-matching checks in place. The UK decisions above date from 2023, and some of those universities may have reviewed them since. The pattern is what matters. A licensed tool can be available and still deliberately unused. ## What a detector score can and cannot count as This section matters more than any other. A detector score is a probability estimate from a model that has an error rate. It is not a record of how the text was written. It cannot show a prompt, a draft history or a deleted paragraph. All it can say is that the writing resembles patterns the model associates with AI output. At most, a score can count as a reason to look more closely and to ask the student about how they produced the work. It cannot count as proof on its own. A fair process gives more weight to other evidence: drafting history, whether the student can talk through the work, and how it compares with their earlier writing. Some writers carry more risk than others. Detectors can misread the work of people writing in English as an additional language, because plain, regular phrasing can look statistically like machine output. The detail is in our piece on [AI detectors and non-native English writers](https://scriptgrain.com/blog/ai-detectors-and-non-native-english-writers). ## How to find out what your own university uses Don't rely on a search result, this one included. Go to the source: 1. **Search your university's website** for "academic integrity" or "generative AI". The central policy usually lives there, separate from course handbooks. 2. **Read your programme's assessment guidance.** Departments sometimes set rules that are stricter or looser than the central policy. 3. **Check your VLE's Turnitin help pages.** They usually say whether the AI indicator is enabled and whether students can see it. Many institutions show it to staff only. 4. **Ask.** Your module leader, personal tutor or academic integrity office can tell you directly. Asking is a reasonable question and says nothing about your own work. If you are a member of staff, add one step: find out who owns the decision to enable the feature, and what the agreed process is when a score is high. ## If a flag is wrong: evidence that helps A false flag is stressful. The best response is calm and documented. You need evidence of how you worked, so don't build your case around arguing with the detector. Gather drafts, version history, notes, reading lists and browser history: anything that shows the work developing over time. If you write in Word, version history can be a strong record, and [how to prove you didn't use AI in Word](https://scriptgrain.com/blog/how-to-prove-you-didnt-use-ai-in-word) walks through it. For other tools, see [how to prove you didn't use AI](https://scriptgrain.com/blog/how-to-prove-you-didnt-use-ai). Be ready to talk through your own choices: why you structured an argument the way you did, where a source came from, what you cut. That is easy when the work is yours. Keep the tone steady. You are not trying to beat a machine. You are showing a person how the piece came to exist. ## What detectors measure, and why that matters for your own writing Detectors look at surface patterns: how predictable the word choices are, how uniform the sentence lengths are, how neatly the structure fits a template. Smooth, even, generic text tends to score higher. Text with more variety and a recognisable personal voice tends to score lower. That is no reason to write badly. It is a reason to know your own habits. If you naturally write in an even, careful register, you may be flagged more often than someone who writes in bursts. If you want to see how your own writing compares with those patterns, ScriptGrain's [AI detect tool](https://scriptgrain.com/ai-detect) flags common AI writing patterns, and the Free plan needs no card. It is no substitute for Turnitin, it cannot predict what your university's tool will say, and its score cannot prove who wrote something. Use it to understand your writing, and use your drafts and version history as evidence. ## Questions ### Do UK universities use AI detectors? Some do, and Turnitin's AI indicator is the most likely tool because it comes with software universities already license. Others opted out: the University of Manchester and Bournemouth University both published decisions in April 2023 not to use Turnitin's first release. Policies change, so check your own institution's current academic integrity page. ### Is Turnitin's AI detection accurate enough to be used as proof? Not on its own. A score is a probability estimate, and it cannot show how a text was written. Even a small error rate adds up to many wrongly flagged students across thousands of submissions. Vanderbilt worked out about 750 possible false flags a year from its own volume. A score can justify a closer look and a conversation. It cannot settle the question. ### Can I see my own AI score before I submit? It depends on how your university has set up Turnitin, and many set it up so that only staff see the AI report. Ask your module leader or check your university's Turnitin help page. Running your work through a random online detector will not tell you what your university's tool will say. ### Why do some universities turn AI detection off? The reasons published by Vanderbilt, Manchester and Bournemouth include the risk of false positives, possible bias against non-native English writers, the lack of any explanation of how the tool works, and too little time to test it and write guidance. Turning detection off does not mean ignoring AI misuse. Integrity rules still apply. ### Do universities use AI detectors for admissions? For UK undergraduate applications, UCAS runs similarity detection on personal statements and tells universities when it finds matches. That compares your text with other statements, which is different from an AI detector. UCAS says submitting an AI-generated statement as your own could be treated as cheating, and applicants must declare the statement was not provided by another source, including AI. ## More from the Journal - [What AI Detector Does Turnitin Use? Its Own Model and Score, Explained](https://scriptgrain.com/blog/what-ai-detector-does-turnitin-use) - [Is the JustDone AI Detector Accurate? What the Tests Show](https://scriptgrain.com/blog/is-justdone-ai-detector-accurate) - [One API key, many client voices: a multi-profile setup for agencies and ghostwriters](https://scriptgrain.com/blog/one-api-key-many-client-voices) [More from the ScriptGrain Journal](https://scriptgrain.com/blog)