What AI detectors do universities use? 17 policies checked

Universities checked

17, in 4 countries

Detector off or never enabled

11

On, with conditions

4 (Turnitin)

Licensed an alternative

2 (Copyleaks)

Common condition

a score is not evidence on its own

Pages read

20 October 2026

The short answer

Most universities that already license Turnitin for plagiarism checking also had access to its AI writing detector, because it arrived inside the same product in April 2023. Many then switched it off or never turned it on. Those that keep it on restrict who sees the score and say in writing that a score alone cannot support a misconduct finding. A smaller group license Copyleaks, which offers both similarity and AI checks.

The 17 policies

Why universities switched detectors off

The reasons on the pages repeat. False positives: Vanderbilt worked through the vendor's claimed 1 percent false positive rate against the 75,000 papers it submitted in 2022 and estimated that about 750 could have been wrongly labelled. Johns Hopkins cites reports of false positives and the fear of falsely accusing students. No explanation of how it works: Vanderbilt notes that the tool was switched on for customers with less than a day's notice and without detail on how it reaches a score. Bias: several pages point to research showing detectors flag non-native English writing more often, the same finding set out in do AI detectors work.

How universities that keep a detector use it

• Staff or instructor access only, so students do not see a raw score (Melbourne, Purdue).
• Use is optional and left to professional judgment (UNSW).
• A report is a reason to look further, never sufficient evidence for an allegation on its own (Melbourne, Purdue, Illinois Springfield).
• Guidance on the tool's limits is a condition of switching it on (Ulster).

The vendor's own documentation supports caution at low scores. In May 2023 Turnitin stated that its document false positive rate is under 1 percent for documents with more than 20 percent AI writing, and added an asterisk to scores under 20 percent. Its current guide says that between 1 and 19 percent no exact figure or highlights are shown, because false positives are more likely in that range. What a percentage does and does not mean is covered in what an AI detector score means.

If your university's detector flagged your work

Read your university's own policy page first: many say a detector report cannot be the only evidence. Then gather evidence of how you wrote the work, as set out in proving you wrote it, and if English is not your first language, say so and point to the published research on false positives for second-language writers. The base-rate arithmetic in false positives and base rates explains why even a low error rate produces many wrong flags across a whole university.

How we built this table

We included only universities that state their decision on a page they publish themselves: teaching centre, IT help, academic integrity office or official news. Secondary lists and blogs were used to find leads and never as evidence; universities whose decision we could not confirm on their own site were left out. Each row was read on 20 October 2026. Our own detector tests follow a separate, published protocol: see how we test. If your university has published a decision that is missing or out of date here, tell us through the contact page with the link and we will check it and update the table.

This guide is part of the testing AI systems hub. It is best read alongside do ai detectors work? what the published tests found and what an ai detector score means for a writer, teacher or editor, which cover the neighbouring questions.