Teachers: what Turnitin AI score do you actually investigate?
Genuinely curious what score other educators use as their cutoff for further investigation. I have been using 40% as my threshold but wondering if that is too high or too low.
Context: I teach college freshman composition. About 180 students per semester across 6 sections.
At 40% threshold I flagged maybe 12 students last semester. 8 admitted to AI use. 4 maintained they wrote it themselves (and I believe 2 of them based on their writing portfolios).
Is there a best practice here or are we all just guessing?
7 Replies
Join the discussion.
Log In to Reply
I switched from GPTZero to Originality AI last month and the difference is significant. Originality AI handles longer texts much better and the output reads more naturally imo.
My professor uses Turnitin and I've been self-checking my work before submitting. The scores are pretty inconsistent though, sometimes it flags perfectly human text at 30%+ which is frustrating.
My professor uses Originality AI and I've been self-checking my work before submitting. The scores are pretty inconsistent though, sometimes it flags perfectly human text at 30%+ which is frustrating.
I've been using Copyleaks for about 3 months now and the results have definitely gotten worse recently. I think the detectors are catching up faster than the humanizers can adapt.