Universities that turned off Turnitin's AI detection
Probably not, unless you are at one of the few dozen institutions that said so in public. Turnitin's AI writing detection is a setting each university controls, so it can be running at your campus and switched off at the one across the city, and neither will necessarily announce which. What is checkable is that a specific set of institutions has turned it off and published the reasons, from Vanderbilt in August 2023 through Washington State University in February 2026, and the reasons repeat almost word for word. Every row in the table below links to that institution's own page. None of it changes what your own course policy permits.
The universities that turned it off
One rule decides what appears below. An institution enters this table only if we opened a page published by that institution saying so: an announcement, a policy, a teaching centre post, or a document under its own name. Nothing here comes from another site's list. That rule is why this table is shorter than the ones you will find elsewhere, and it is also why you can cite it.
Two things are worth knowing before you read it. AI writing detection is a separate feature from Turnitin's similarity or text-matching report, and almost every institution below switched off the first while keeping the second running. And a university turning the feature off changes what evidence an instructor can bring to a case. It does not change what your course allows you to do.
| Institution | Country | What was turned off | Date | Source |
|---|---|---|---|---|
| University of Arizona | USA | Turnitin AI Detection "temporarily disabled for all UArizona courses", on the grounds that the feature "generated inconsistent results" | 27 June 2023 | Dean of Students |
| University of Pittsburgh | USA | "the Teaching Center has disabled the AI-detection tool in Turnitin", and endorses no detector of any kind | 2023, page reviewed 9 February 2026 | Teaching Center |
| Vanderbilt University | USA | Turnitin's AI detector disabled for the foreseeable future; text matching kept | 16 August 2023 | Brightspace news |
| Georgetown University | USA | AI writing detection turned off in similarity checks, on concerns raised by the main campus faculty executive committee and the Honor Council | October 2023 | Georgetown UIS |
| Australian National University | Australia | "ANU has disabled this functionality as of 1 January, 2024 in Turnitin and it is no longer available for ANU staff." | 1 January 2024 | TELT blog |
| UC Berkeley | USA | Opted out campuswide after an opt-in pilot that ran from Fall 2023 to the start of Spring 2025 | Post dated 1 May 2025 | RTL Commons |
| University of Waterloo | Canada | AI detection functionality in Turnitin discontinued after consultation with associate deans, the Undergraduate Operations Committee and the EdTech Steering Committee | September 2025 | AVP Academic |
| UCLA (HumTech) | USA | Declined to adopt Turnitin's AI detection, citing concerns and unanswered questions about accuracy. A unit-level post, not a provost policy. | 9 October 2025 | HumTech |
| Curtin University | Australia | "Only the Gen-AI detection feature of Turnitin will be disabled, the regular text-matching originality checks will remain active." | Announced 4 September 2025, effective 1 January 2026 | OASIS news |
| Indiana University | USA | "No AI detection tools (such as GPTZero, Turnitin AI Detection, or Originality.AI) are approved for use at Kelley or Indiana University." Turnitin is licensed for plagiarism review only. | Teaching page updated 12 January 2026 | CITL, Kelley playbook |
| Washington State University | USA | Contract for Turnitin's AI Detection cancelled. The Turnitin plagiarism package stays available in Canvas. | Memo dated 11 February 2026 | Provost memo, policy page |
| Yale University | USA | "The AI detection feature of Turnitin is currently disabled." | Not dated | Poorvu Center |
| Michigan State University | USA | "Turnitin AI Detection is not currently available at MSU" | Not dated | D2L help |
| New York University | USA | "AI Writing Detection has been disabled in the Turnitin Similarity Reports." | Not dated, guide updated 10 June 2026 | Libraries guide |
| CU Boulder | USA | "The Turnitin AI Writing Detector, a feature that detects AI writing in originality reports, is not currently available at CU Boulder." The page adds that the decision can be reconsidered in the future. | Not dated | CU Boulder OIT |
| Colorado State University | USA | A pause rather than a ban. Its teaching institute recommended pausing the rollout, and explains why staff cannot find the tool. | Not dated | TILT |
| MIT Sloan | USA | Recommends against detectors: "AI detection software is far from foolproof". A recommendation, not a confirmed disable. | Not dated | Sloan EdTech |
Six of those rows say "not dated", and that is deliberate. Yale, Michigan State, NYU, CU Boulder, Colorado State and MIT Sloan each state a current position on a page that carries no decision date. We are not going to invent one for them. It also means those are the six rows most likely to go stale without warning, because a sentence with no date attached can be edited without anyone announcing it.
Notice how carefully several of them hedge. Arizona wrote "temporarily". Colorado State called it a pause. CU Boulder wrote that the decision can be reconsidered. None of those is a permanent settlement, and reading them as one would be reading more into the page than the page says.
The ones we could not verify
A handful of universities appear in almost every list on this subject, and they are not in the table above. Several of them have probably done what is claimed. We could not open a page from the institution itself to confirm it, so they are here instead.
The reason for the split is simple. Saying that a named university changed its academic integrity practice is a claim about a real organisation that students and staff will act on. A claim like that either has a source or it does not, and a chain of blogs quoting each other is not a source. This list is short on purpose, and rows move up to the table above as soon as a primary turns up.
| Institution | What is claimed, and by whom | Why it is not in the table above |
|---|---|---|
| Northwestern | Inside Higher Ed reported on 5 August 2026 that it has disabled Turnitin's AI detection. Every roundup repeats it. | We found no Northwestern page saying so. Its teaching centre and IT service URLs either return errors or redirect elsewhere. The Inside Higher Ed sentence is the outlet's own reporting rather than a statement issued by the university. |
| Johns Hopkins | Named everywhere. Roundups describe an "advisory only" policy dating from early 2025. | The university teaching page that would carry it refused our request. The specific wording quoted around the web appears only in aggregator prose, never with a link to the source document. |
| University of Washington | Named in Washington State University's own provost memo as an R1 institution that has banned the tool's use in academic integrity. | That is one university describing another, which is not the same as its own announcement. UW's Turnitin service page is gone, and its teaching and library AI guidance says nothing about detection. |
| Oregon State University | Also named in the WSU provost memo. | Its student conduct and AI guideline pages both refused our requests, and its Turnitin help page for faculty does not mention AI detection either way. |
| Australian Catholic University | Listed as March 2025 by several roundups. | No institutional source. The March 2025 date appears without attribution wherever it is printed. |
| University of Cape Town | Listed as October 2025, and it is the entire basis for the "five countries" phrasing used elsewhere. | The teaching centre page that would carry it does not exist at the address given. No primary found. |
| Western University | The Western Gazette, the student newspaper, reported that it stopped using Turnitin's AI detection. | The article would not load for us. This is likely a good row once we can read it, and it is first in the queue at the next check. |
| Alabama, SMU, Iowa State | All three trace to a single social media post from December 2023. | A post is not an announcement, and we found no institutional page for any of the three. |
| About thirty more | Baylor, Boston University, DePaul, Syracuse, UC Irvine, Maryland, Toronto, UBC, Simon Fraser, Deakin, Macquarie, Dundee, Edinburgh, Manchester, Nottingham and others, all via one registry and its copies. | Not independently checked yet. Several are probably true. They join the table one at a time, each when its own source is confirmed. |
One trap is worth naming, because at least one competing page has fallen into it. Searching for the Washington State decision surfaces a WSU announcement headlined "Turnitin contract cancelled; faculty encouraged to use SafeAssign", which describes WSU dropping Turnitin entirely and moving to a different tool. That announcement is dated August 2016. It is a real page about a real decision, a decade before the one on this list, and it is not evidence of anything that happened in 2026.
The reasons they gave
Read enough of these announcements and the same four arguments come back, usually in the same order.
The arithmetic of a small percentage. Vanderbilt did the multiplication in public: "Vanderbilt submitted 75,000 papers to Turnitin in 2022. If this AI detection tool was available then, around 750 student papers could have been incorrectly labeled as having some of it written by AI." Washington State repeated the exercise with its own volume, reporting that Turnitin analysed 148,547 assessments there in Fall 2024, so even a 1% false positive rate implies almost 1,485 assessments wrongly flagged.
It does not err evenly. Vanderbilt noted that "AI detectors have been found to be more likely to label text written by non-native English speakers as AI-written". Waterloo put it the same way, that the tools "have also been found to be biased toward students whose first language is not English", and Washington State extended the concern to neurodivergent students as well.
A percentage is not evidence. This is the argument that seems to decide it for most institutions. UC Berkeley wrote that unlike a similarity match, "the AI detection tool provides no such conclusive proof, and it is unclear how these findings would or should be used in an academic integrity investigation." Washington State's policy is blunter: suspicion of AI use is not sufficient for a finding of student responsibility, and no detector may be the sole support for a case.
The student never sees it. Washington State makes one argument nobody else does, and it is the most interesting of the four. The detector's output is instructor facing rather than student facing, which means a student has no way to raise a concern about their own work before the report reaches the marker. There is no moment where the person most affected can say anything.
Turnitin's own chief product officer, Annie Chechitelli, told Inside Higher Ed in June 2023 that "the sentence-level false positive rate is approximately 4 percent", and that when the tool reports a document as having a less than 20 percent chance of machine writing, "it has a higher incidence of false positives", so the company added an asterisk to those results. An instructor acting on a 15 percent score is acting on a number the vendor itself marks as unreliable.
Waterloo added a reason the others leave out, which is that the feature costs money. Its own technical group found the advantages inconclusive and reported that "in more than one instance the product flagged human written text as 100% generated by AI, which is concerning", and the university concluded the cost outweighed the benefit. None of these institutions said that unauthorised AI use had stopped mattering. They said this particular instrument could not measure it well enough to act on.
What the "more than 50 universities" figure counts
You will see a number in the headline of almost every page on this topic. It is usually "more than 50 universities across five countries", sometimes 60. It is not invented, but it does not mean what the headline implies, and it is worth following back.
The pages that print it cite a tracker. That tracker cites two sources and provides no institutional links of its own. One of those two is more honest than everything downstream of it: it calls its own figure documented and growing rather than precise, states plainly that no authoritative census exists, and recommends checking each school's own page. The other is the root of the whole chain, a registry maintained by the nonprofit PLEASE.
That registry is a real piece of work and better sourced than any commercial page copying it, because it links a source per row. Three things about it do not survive the trip into a headline. It was last updated in September 2024, so it cannot possibly include Waterloo, Curtin, UC Berkeley or Washington State. It carries no dates on individual entries. And it deliberately separates institutions into three groups, "banned", "recommend against", and "AI detection disabled with no further public statement", which the copies flatten into one number. The "five countries" phrasing is drift too. The registry lists four.
The defensible statement is narrower and more useful. Inside Higher Ed reported on 5 August 2026 that at least a dozen universities have disabled Turnitin's AI detection. The PLEASE registry lists roughly 57 institutions across four countries, on a page that has not been updated since September 2024. This article verifies 17, because it requires each one's own source, and that is the whole difference between the numbers.
The universities that kept it
Start with the one that goes the other way. NC State's teaching and learning group announced on 2 September 2025 that "Turnitin AI Detection is re-enabled and available through the Turnitin Assignment tool", after a period without it. The same post tells instructors that "AI detection should start conversations about responsible use, as results can include false positives". A list that only runs in one direction would be an argument dressed as a reference, and this row is the reason this one is not.
Plenty of institutions never switched it off. UW-Whitewater's teaching centre published guidance in January 2026 that keeps the tool and puts rules around it: "Treat the AI score as a conversation starter, not a verdict", and "AI detection should never be the sole evidence in a misconduct case." The University of Kentucky's provost office wrote to faculty back in April 2023 that there was no way to opt out of seeing the score, and that scores should be handled sensitively and with healthy skepticism.
The pattern across both camps is more consistent than the headlines suggest. Almost nobody defends the score as proof. The universities that kept the tool and the universities that dropped it broadly agree on what it is worth on its own, which is not much, and disagree about whether something worth that little is worth having in front of instructors at all.
What is replacing it, where anything is, tends to be assessment design rather than a different detector. Washington State's memo points instructors at the University of Sydney's two-lane framework, published in March 2025, which sorts assessments into secure supervised work on one side and open work where AI use is expected and integrated on the other. Indiana's guidance for faculty is the most concrete version we found: skip the detector, look at the work you already have, talk with the student from curiosity rather than accusation, ask them to explain or adapt the piece, document facts rather than conclusions, and then use the ordinary misconduct process.
What the Stanford study actually measured
One paper does most of the work in these announcements, and it is worth reading rather than inheriting. Liang, Yuksekgonul, Mao, Wu and Zou published GPT detectors are biased against non-native English writers in Patterns on 10 July 2023. They ran seven publicly available detectors over 91 TOEFL essays written by non-native English speakers and 88 essays by US eighth-graders.
The detectors were close to perfect on the eighth-grade essays. On the TOEFL essays the average false positive rate was 61.22%. The figure people usually quote next is 97.80%, and it means something narrower than it sounds: 89 of the 91 essays were flagged by at least one of the seven detectors. The number for unanimous agreement is much smaller, 18 essays out of 91. The authors then asked a language model to enrich the vocabulary in the same TOEFL essays, and the average false positive rate fell from 61.22% to 11.77%, which is the finding that gives the mechanism away. These tools are reacting to plainer and more predictable prose, not to how the text was produced.
Two corrections, because both are common. The circulating figure of 61.3% is wrong, and the paper says 61.22%. More importantly, Turnitin was not one of the seven detectors tested. The word does not appear in the paper at all, and the detectors were accessed in March 2023, weeks before Turnitin's own tool launched. Institutions cite this study as evidence about the class of detectors and the statistical property they share, and that is a fair use of it, but Turnitin's tool was not among those measured. Anyone telling you the study proves something specific about your Turnitin score has not read it.
How to find out about your own university
No list can answer this for you, including this one, because most institutions never announced their setting either way. Four places are worth checking, roughly in this order: your teaching and learning centre's website, the LMS help documentation for Turnitin, the academic integrity or student conduct pages, and any announcements from the provost or academic senate. Almost every institution above documented its decision in exactly one of those four and nowhere else.
If none of them mention AI detection, the setting is simply not public, and the answer is a short email to the academic integrity office asking whether the AI writing indicator is enabled for your courses. That is a reasonable question and staff answer it.
It is worth being clear about what a yes or a no gets you. If the detector is off at your university, one weak instrument has been removed from a process that still exists. Curtin's own wording is that all assessments continue to be reviewed for academic integrity. Washington State still runs misconduct cases, just not on a score alone. Indiana tells faculty to look at drafts, checkpoints and in-class work instead. If your course requires you to disclose AI use, that obligation is untouched by a setting in a piece of software, and whether a particular use counts as cheating was never a detection question in the first place.
If the detector is on, the useful reading is what the score actually reports and what tends to raise it on genuinely human work. If a score has already turned into an allegation, we wrote separately about what to do when you are accused of using AI on an essay. And the risk that outlasts every detector policy on this page is a citation that does not check out, because a reference nobody can find is a problem no institutional setting makes go away.