@prosewitch False positive/false negative rates are currently somewhere between 5% and over 50%. If you do lots & lots of tests, most (almost all) of the AI texts flagged as AI will be wrong, because at the moment, most text is human generated.
Comparison to a known person's normal writing is probably a better test.
Ref:
https://www.washingtonpost.com/technology/2023/06/02/turnitin-ai-cheating-detector-accuracy/