Do universities use AI detectors?

Most articles answering this question are written by companies that sell AI detectors. We did something simpler: we picked 300 well-known universities before looking anything up, then read each one's own teaching-centre, IT or academic-integrity pages to see what it says about AI-writing detection. Every school we name below links to the page we read.

Coursework only. This covers essays and assignments marked by instructors. Admissions offices rarely publish whether they screen application essays, so this data says nothing about admissions.

The short answer, in numbers

• Not in use: 100 of 300 (33%). Of the 249 universities with a public policy, 40%.
• Not in use, or allowed only with cautions: 196 of 300 (65%); 79% of those with a policy.
• Which detector schools provide: of the 53 that provide one, 44 use Turnitin's, 4 Copyleaks, 1 Pangram, and 4 don't name the product. Turnitin is also the tool most policies mention: 193 of 249 policy pages name it.

"No public policy" means we searched the university's site and found no page on AI detection. When we searched those schools a second time, about one in five turned out to have a policy after all, so we lean on the shares among schools that publish one.

Elite universities are the least likely to run one

The split by type of school is the clearest pattern in the data. Half of the 50 best-known US universities don't run an AI detector, and only 3 provide one. Among large community colleges, none of the 19 with a public policy has switched detection off or declined it, and 12 of the 31 provide one.

* Most liberal-arts and online universities we checked publish no policy on AI detection, so their counts are too small to compare.

By country

We checked 10 UK and 20 each of Canadian and Australian universities, against 250 in the US, so read the non-US rows as counts, not national rates. Six of the 10 UK universities and 11 of the 20 Canadian ones don't run a detector; Australia is the most split, with 8 not running one and 6 providing one.

Universities are still switching it off

About half of the switch-offs date to 2023, the year Turnitin added AI detection and many schools opted out of the preview. But the decisions haven't stopped. Universities whose pages give a 2025 or 2026 decision date:

• UC San Diego (Extended Studies): deactivated April 7, 2025.
• UC Berkeley: ended its pilot in spring 2025.
• University of Queensland: turned off in mid 2025.
• University of Waterloo: discontinued from September 2025.
• Curtin University: turned off from January 1, 2026.
• Washington State University: cancelled the AI-detection contract in February 2026, keeping plagiarism checking.

All 36 switch-offs, with the date as each page gives it:

"Date not given" means the page says the detector is off but not when the decision was made.

Every university we checked

Search by school name or short name (MIT, UCLA, NYU) with country and detector filters, or filter the table here. For the 48 schools students search for most by name (NYU, ASU, U of T, UCL and more), we re-read each school's page in October and wrote a short answer.

247 universities with a public policy. Type a name to find yours. The 51 schools with no public page are counted above but not listed, and 2 more are counted but not listed because their page couldn't be opened directly to confirm it.

What this means if you're a student

• Check your own school's page, then your course. A university-wide decision isn't the last word: where a detector is allowed, individual instructors usually decide whether to use it, and some departments set their own rules.
• "Not in use" doesn't mean AI is allowed. A school's AI rules sit in its academic-integrity policy and your course syllabus, whether or not it runs a detector.
• Where it is on, you won't see the score. Turnitin's AI score is shown to instructors, not students. More in does Turnitin detect AI humanizers?
• Keep your drafts. Whatever your school uses, your writing history is the best evidence you wrote your work. How often detectors wrongly flag human writing, study by study: AI detector false-positive rates.

How we built this

• The list. 300 universities, fixed before any policy was looked up: 50 prominent US universities (the Ivy League and peers, plus large public flagships), 200 further US institutions (other universities, liberal-arts colleges, online universities and large community colleges), and the 10 leading UK, 20 Canadian and 20 Australian universities. It is an editorial pick of large, well-known schools, not a random sample, so the shares describe these 300, not every college.
• The source. Only the university's own pages count: teaching centre, IT service, academic-integrity office or provost guidance. News reports were used only to find those pages.
• Checks. We re-opened a sample of rows against their pages (no status errors found), and searched again for every school where we'd first found no policy. Where a page names a single college or department rather than the whole university, the table says so.
• Dates. All pages were read on Oct 9, 2026 (US Eastern). Policies change; we plan to re-check this list every month. Spotted a change at your school? Email hello@humanizernerd.com with the link.
• Correction, Oct 23, 2026. Dartmouth moved from "Provided" to "Allowed, with cautions". Its September 2026 guidelines authorize faculty to use Pangram, but say the college has not decided whether to buy an institutional subscription. Totals changed by one: 53 provide a detector (was 54), 96 allow one with cautions (was 95).

The full table is free to reuse with a link to this page (CC BY 4.0): download the CSV.