> Fact is that I maybe saw it in 10% of blogs and news articles before Chatgpt.

I believe you. But also be aware of the Frequency Illusion. The fact that someone mentions that as an LLM signal also makes you see it more.

https://en.wikipedia.org/wiki/Frequency_illusion

> Yes it's not a guarantee but it is at least a very good signal that something was at least partially LLM written.

Which is perfectly congruent with what I said with emphasis:

> it is never sufficient on its own to identify LLM use

I have no quarrel with using it as one signal. My beef is when it’s used as the principal or sole signal.