That tool (and any of the other ones that currently exist) aren’t accurate. They say things are AI that aren’t and say things are real that are AI.
I’ve tested them on articles generated with davinci, ChatGPT, and personally written. Results are widely inaccurate.
only way Google could possibly detect AI without punishing everyone is if OpenAI adds a watermark. Then at that point the rephrasing tools would be used and that removes that as an answer as well.
My .02 cents Google will go heavy into user metrics such as bounces back to SERP and reading other results as well as time spent on the page. That won’t hurt AI arricles though just crappy AI articles that are auto generated.
Oh and if you play around with ChatGPT and davinci’s prompts you can get articles that look much more human than stuff that I’ve actually written!