What AI detector do universities use? What 19 universities say in 2026
What AI detector do universities use? When they use one at all, it is almost always Turnitin's AI writing indicator, because it comes with the plagiarism checker they already license. But many universities have switched it off or never turned it on: Curtin University stopped using it on 1 January 2026, Waterloo in September 2025, and Edinburgh's is switched off, while Yale, Glasgow and CU Boulder support no AI detector at all. Here is what 19 universities say on their own pages, and how the ones that keep a detector treat its score.
The short answer: mostly Turnitin, and only staff see it
Turnitin sells its AI writing detection to institutions that license Turnitin Originality, or as an add-on to iThenticate 2.0, its product for research and publishing. It does not sell individual licences. Only instructors and administrators can see the indicator; students can't, though an instructor can download the report as a PDF and share it.
The indicator needs at least 300 words of prose and takes files of up to 30,000 words. It covers English, Spanish, Japanese, Arabic, Dutch, German and Turkish, and it leaves out lists, headers, references, data tables, formulas and code. Scores from 1% to 19% appear only as an asterisk. Turnitin itself says the percentage "should not be used as the sole basis for action". More on how it works is in does Turnitin detect AI?
The only other tool we found named on a university's own page was Copyleaks, which Utah State University said in 2023 it was using in place of Turnitin. We found no official page naming GPTZero or Pangram as an institutionally licensed detector.
Universities that use an AI detector
| University | Tool | How they say to use it |
|---|---|---|
| University of Melbourne (Australia) | Turnitin | Staff only. A high score alone is not grounds for an allegation; a second piece of evidence is needed |
| University of Bristol (UK) | Turnitin | Instructors only, and "only when concerns arise"; staff accept a disclaimer before scores appear |
| University of Canterbury (New Zealand) | Turnitin, in some courses | A first step to find work that needs a closer look, not generally the sole evidence (September 2026) |
| Maastricht University (Netherlands) | Turnitin | Papers sent for a plagiarism check are checked for AI too; Dutch papers since 6 October 2026 |
| NC State University (US) | Turnitin | Re-enabled in September 2025, with a warning that results can include false positives |
| University of Regina (Canada) | Turnitin | The only approved detector; outputs are never definitive proof (February 2025) |
| Utah State University (US) | Copyleaks | Used in place of Turnitin since July 2023, including AI checks |
Universities that switched detectors off or don't use one
| University | Decision | Reason given |
|---|---|---|
| Curtin University (Australia) | Turnitin's AI detection disabled from 1 January 2026 | "Trust and clarity within a modern academic culture" |
| University of Waterloo (Canada) | Turnitin's AI function discontinued, September 2025 | Unreliable and biased against students whose first language isn't English; internal tests flagged human writing as "100% generated by AI" |
| University of Edinburgh (UK) | Submissions not processed for AI writing detection (September 2026) | Switched off by the service owner; no reason given |
| University of Auckland (New Zealand) | Asked Turnitin not to enable the feature after a pilot | Concerns about a lack of testing and transparency |
| University of Glasgow (UK) | Does not use AI detection tools (2026 guidance) | Investigations should never rely on such software; staff must not upload student work to outside detectors |
| Yale University (US) | Teaching centre doesn't endorse AI detection or enable it in Canvas | Cites research calling detectors unsuitable for high-stakes use |
| University of Colorado Boulder (US) | Supports no AI detection tool | Inaccurate results and privacy |
| University of East Anglia (UK) | Told a freedom-of-information request in November 2025 that it uses no AI detection software | None given |
| Vanderbilt University (US) | Turnitin's AI detector disabled, August 2023 | At a 1% error rate, about 750 of the 75,000 papers it submitted in 2022 could have been wrongly flagged; no insight into how it works |
| University of Tartu (Estonia) | Using AI detection software is "not justified" | Not technically possible to identify AI text reliably |
One caveat on CU Boulder: its IT office says it is re-examining Turnitin's indicator, with no date or outcome given.
What Swiss universities say
ETH Zurich's academic integrity page says technical recognition of AI output "is currently unreliable and will probably remain so", so trusting it is "not appropriate". EPFL's guidelines from April 2025 say detectors such as Turnitin are not admissible as stand-alone evidence of AI plagiarism, because of their high risk of false positives. Neither page says whether the university licenses a detector.
Why universities switch AI detectors off
The reasons repeat from one announcement to the next:
- False positives at scale. Vanderbilt did the sums: even a 1% error rate would have meant about 750 wrongly flagged papers in 2022 alone.
- Bias against second-language writers. Waterloo and Vanderbilt both cite it, a concern that goes back to a 2023 Stanford study (see AI detectors and non-native English).
- No way to explain a score. Vanderbilt and Auckland point to a lack of testing and transparency.
- Privacy. CU Boulder names it, and Glasgow bars staff from uploading student work to outside detectors.
Recent research backs the caution on academic writing. In a June 2026 study of 160 long academic papers, Turnitin scored all 40 fully AI-written papers below 20% while flagging no human ones, and the authors concluded that detectors "should not be used as sole evidence in high-stakes decision-making". A Notre Dame study of 642 research abstracts found light AI editing was flagged far more often than text run through a humanizer. Our round-up of detector studies has the details.
How universities that keep a detector use the score
The universities that still run Turnitin set similar limits. The score goes to staff, not students. It is a reason to look closer, not a finding: Melbourne requires a second piece of evidence, Regina says outputs are never definitive proof, Canterbury calls it an initial step, and Bristol asks staff to use it only when they already have concerns. Melbourne also tells staff to focus on work where more than 20% is flagged.
What this means for students
You can't run your work through your university's Turnitin, and you won't see the AI score unless your instructor shares it. What protects you is the same at every university: know the AI rules for each course, keep drafts and version history, and be ready to talk through your sources and argument. If you are flagged for work you wrote yourself, our guide to being wrongly flagged by an AI detector explains what to do.
For a private look at your own draft before you submit it, IsThisAiWritten is free, needs no account and has no word limit, so a whole dissertation goes in at once. It runs in your browser, so your work is never uploaded, stored or used for training, and its sentence-by-sentence highlights show which passages read as the most AI-like.
How we checked
We used only universities' own pages (policy, teaching centre, IT and integrity pages), all opened on 1 November 2026, plus Turnitin's guides. Dates are the decision date or the page date where one is given. We left out pages we couldn't open and decisions that later pages had overtaken. Policies change, so check your own university's current page before relying on any of this.
Sources
- Turnitin, AI writing detection capabilities FAQs, checked November 2026
- Turnitin, Can I purchase an individual Turnitin license?, checked November 2026
- University of Melbourne, Turnitin's AI writing detection tool, checked November 2026
- University of Bristol, Turnitin AI writing detection, checked November 2026
- University of Bristol, New Turnitin AI writing score column (8 August 2025, updated 22 January 2026), checked November 2026
- University of Canterbury, Gen AI FAQs for students, checked November 2026
- Maastricht University Library, Turnitin FAQ, checked November 2026
- NC State DELTA, Turnitin updates: AI detection and more (2 September 2025), checked November 2026
- University of Regina, AI-writing detection tools (revised 14 February 2025), checked November 2026
- Utah State University, Turnitin overview, checked November 2026
- Curtin University, Update on Turnitin AI detection tool (4 September 2025), checked November 2026
- University of Waterloo, Discontinuing use of AI detection functionality in Turnitin, checked November 2026
- University of Edinburgh, Turnitin (25 September 2026), checked November 2026
- University of Auckland, Turnitin (updated 1 July 2026), checked November 2026
- University of Glasgow, GenAI guidance: identifying potential concerns, checked November 2026
- Yale Poorvu Center, AI guidance for teachers: academic integrity, checked November 2026
- CU Boulder Center for Teaching & Learning, AI ethics, checked November 2026
- CU Boulder OIT, AI writing detection, checked November 2026
- University of East Anglia, FOI 25-334: AI detection software (5 November 2025), checked November 2026
- Vanderbilt University, Guidance on AI detection and why we're disabling Turnitin's AI detector (16 August 2023), checked November 2026
- University of Tartu, Guidelines for using AI applications in teaching and studies, checked November 2026
- ETH Zurich, AI in education: academic integrity, checked November 2026
- EPFL, Guidelines for the use of AI in teaching (April 2025), checked November 2026
- Van Vlasselaer, Van Droogenbroeck and Spruyt, Who wrote this? (International Journal for Educational Integrity, 2026), checked November 2026
- Karr, Khvatskii, Hua and Chawla, Why AI Detection Fails for Academic Integrity (2026), checked November 2026
- Pangram, Does Pangram integrate with my LMS?, checked November 2026
- GPTZero pricing, checked November 2026
Questions
Do universities use GPTZero or Pangram?
We found no official university page naming GPTZero or Pangram as an institutionally licensed detector. Both sell to educators: Pangram connects to Canvas, Moodle, Blackboard, Brightspace, Schoology and Google Classroom, and GPTZero's Professional plan connects to Canvas and Google Classroom.
Do universities check essays for AI watermarks?
We found no university page that says so. There is no public checker for text watermarks: Anthropic lists educational organisations among those that can apply for its private detection preview, and OpenAI gives access to approved researchers and expert organisations case by case.
Is this AI written? Check it now.
Free, no sign-up, no word limit. The check runs in your browser, so your text never leaves your device.
Check text free