You can still check if its a real human if you do something really stupid or speak or write giberisch. Almost every AI will try to reply to it or say “Sorry i couldnt understand it” or recent events ( most of the LLMs arent trained on the newest events )