As far as I could tell, for Sapling.ai everything would be AI if there were very few grammatical errors in the text.
IMO, this assumption is wrong.
However, the absence of errors cannot be evidence of the use of AI,
Agreed. I am sure everyone would agree to this while investigating any post for AI.
the use of words that AI loves is becoming a normal phenomenon for most people since most websites and news on the Internet pass their materials through a robot, and we are simply used to using these words more often.
Hmm, interesting, but I still think it is very hard to match the
perplexity (which is the basis of AI detection) of an AI chatbot, especially with so many long and detailed posts.