By skyforbes I used GPTZero to double-check AI-written text, and the results were weird and inaccurate. Paraphrased AI text passed as human, and some fully human text got flagged as AI (likely). There are threads from a year ago flagging how it is inaccurate, but I wanna know if someone else is seeing the same issues still arising as of late 2025.
I find this problematic as many revisers would use it as the sole decider if content is AI or not, and cause many problems to people who wrote stuff themselves (false accusations of cheating)
I find this problematic as many revisers would use it as the sole decider if content is AI or not, and cause many problems to people who wrote stuff themselves (false accusations of cheating)