Do universities use AI detectors?
Widely, automatically, and usually without telling you the result. Which means the score exists whether or not anybody mentions it.
The AI indicator is built into the submission systems universities already use, it runs on submissions by default in most configurations, and the output is visible to staff rather than to students.
So the practical position is this: your work is probably being scored, you will probably never see the number, and you will only hear about it if it is high enough that somebody decides to raise it.
Where it runs
Inside the plagiarism tooling already wired into the learning management system, on the same submission, at the same time, producing a second score on the same report. Nobody has to switch it on per assignment.
Publishers do the same on journal submissions, which catches researchers rather than students. See AI detection and journal submission.
What the number is
A classifier's probability that your writing has properties characteristic of generated text. It matches nothing and finds nothing, which is the difference between it and the similarity score beside it. See how Turnitin AI detection works.
Turnitin publishes the rate at which it is wrong about genuine human writing: 4%.
At any real volume that is a large number of documents. A programme processing 10,000 submissions a term expects hundreds where the indicator is wrong about somebody's own writing.
Ours is 0.4%, which is 99.6% specificity, measured across 15,900 documents of real academic writing.
The part that is not evenly distributed
Wrong flags do not fall randomly across a cohort. They concentrate in particular kinds of writing, and every one of these is common in a university.
Formal, conventional prose. Which is what you are asked to write.
Technical and scientific sections. Methods, procedures, anything formulaic by design.
Heavily revised work. Editing smooths, and smoothing removes the variation a classifier reads as human.
Writing by non-native English speakers. Repeatedly measured at elevated rates in published evaluations, and a large share of any international cohort. See why non-native English writing gets flagged and international students and AI detection.
The published average understates what these groups actually experience.
What is available to you
You cannot opt out and you cannot see the report. You can look at your own document through a detector before you submit it, which is the only point in the process where the information is useful.
That check tells you which paragraphs read flat and gives you the chance to change them. The changes are mechanical and do not touch your argument: vary sentence length, cut connective openings, keep the specific detail. See what makes text read as machine written.
Ours is free and unlimited with no account, takes .pdf, .docx and .tex, has no word cap, scores every paragraph separately, and never stores your text.