The concept of an "AI detector" for written works is absurd. The explicit modelling goal of ChatGPT is to write as humans do and it is extremely good at that.

The concept of an "AI detector" for written works is absurd. The explicit modelling goal of ChatGPT is to write as humans do and it is extremely good at that. ChatGPT if not promted specifically has a very particular tone, which you can surely detect. But writing in that careful, somewhat formulaic english can not be taken as evidence of anything. I can absolutely believe that ESL students tend towards that tone. But that only shows how absolutely useless the detector is.