AI Detectors Don't Work. Teach Instead.

AI detectors promise to tell you whether a student used AI. What they actually deliver is a probability dressed up as a verdict — and the errors don't fall evenly.

Study after study has found these tools disproportionately flag writing by non-native English speakers. The pattern is cruel: the students working hardest to write clearly in a second language are the most likely to be accused of cheating by a machine. A false accusation doesn't just embarrass a student; it teaches them the system is rigged.

Here's the deeper problem: detection is an arms race the school can't win. Every detector update is met with a paraphrasing trick. Meanwhile, teachers spend their energy policing instead of teaching.

There's a better use of that energy. Design assignments where the thinking is visible — outlines, drafts, reflections on what changed between versions. A student who walks you through their process has nothing to hide and everything to show. Conference with students about their work; five minutes of conversation reveals more than any detector score.

Drop the detector. Spend the money on smaller class conferences if you must spend it at all. Trust is a better cheat-prevention system than software, and unlike the software, it doesn't discriminate.

I explain frameworks; I don't give legal counsel.