“can ai tell when something is written by another ai?”

AI detection tools can flag AI-written text sometimes, but they’re not reliable. Turnitin’s AI checker is the one most UK schools use, and it does catch some AI-written work. It also gets things wrong regularly, in both directions. It sometimes flags writing by non-native English speakers and some neurodivergent children as AI-written when it isn’t. And it can miss genuinely AI-written text too, especially once a child has edited it themselves.

There’s another factor worth knowing. AI tools tend to produce similar answers to similar prompts. So if a teacher sets one essay question and several pupils use the same AI tool to help, those essays often come back with overlapping phrasing, structure, or ideas, even if no single one of them looks obviously AI-written on its own.

The practical result is that a detection tool checking one essay in isolation isn’t something to rely on. But a teacher reading a whole set of essays together is often the more reliable check, because the repetition across several answers is its own giveaway.