Try the Turing-test! 😉

It's easy to identify the bots nowadays (let's call them bots but I think you wanted to mention e.g. AI driven LLMs):

- a bot always replies too quickly and precisely

- a bot articulate the sentences too well, very polite

- a bot uses better grammar and rarely uses abbrevs and slangs

- a bot doesn't misspell (like a human - e.g. I have thick fingers and misspelling some words)

- the bot's reply is well summarized

- a bot doesn't swear, only if it was thought previously to do this

- if a bot is lying and you tell this fact to it, it will apologize for the misunderstanding

- a bot doesn't know "everything", try cross-examinate it

- a bot doesn't have feelings, so you can't upset it

So I think 2-3 reply is more than enough to pick if it is a bot. Then you shoud ban/hide it.