Is the Turnitin AI Detector Accurate? Real Classroom Results
I collected some informal data from my own classroom and a few colleagues and wanted to share what we found about Turnitin’s AI detection accuracy in real-world conditions.
Over the course of a semester, we tracked Turnitin AI detection scores against what we knew (through process-based assessment) about whether students used AI. We weren’t doing a formal study, but the pattern was informative.
For papers we were confident were entirely AI-generated (based on student admission or overwhelming evidence), Turnitin flagged them at 80-100% about 85% of the time. The misses tended to be papers where students had used AI humanizer tools or done significant editing after generation.
For papers we were confident were entirely human-written, Turnitin correctly identified them about 96% of the time. But that 4% false positive rate caused real problems. That’s roughly one student per class getting incorrectly flagged each semester. For ESL students specifically, the false positive rate was noticeably higher.
For mixed papers (some AI, some human), results were unpredictable. Scores ranged from 15% to 70% AI for papers we knew had significant AI assistance. The sentence-level detection meant that AI sections scored high while human sections pulled the average down, making it hard to interpret.
The bottom line: Turnitin’s AI detection is a useful screening tool. If a score comes back above 80%, you probably have a genuine case. But scores between 20-60% are too ambiguous to act on without additional evidence. And for ESL students, I’d raise that threshold even higher.
Are you keeping track of accuracy in your classroom? What patterns have you noticed?
4 Replies
Join the discussion.
Log In to Reply
This is EXACTLY what I needed. I've been trying to explain to my department head why we can't just rely on Turnitin scores and now I have actual data to back it up. Sharing this with my whole team tomorrow!
sorry if this is a dumb question but do these detectors work on google docs or do you have to copy paste the text somewhere? just started teaching this year and feeling pretty overwhelmed by all this
im new to all this ai stuff. my school just told us to 'use our judgment' which is super helpful lol. this thread is really helping me understand whats going on tho
I've been teaching for 25 years and I've seen every cheating method imaginable. What concerns me about the AI detection approach is that it shifts the burden of proof. In the past, we needed evidence to accuse a student. Now, a score above some threshold triggers investigation. That's a fundamental change in the teacher-student relationship that I don't think we've fully reckoned with.