There are some systems which can use the LLMs themselves to detect writing (basically, if the text matches what the LLM would predict too well, it's probably LLM generated), but they are far from infallible (with both false positives and false negatives). There's also certain tropes and quirks which LLMs tend to over-use which can be fairly obvious tells but they can be suppressed and they do represent how some people actually write.
Comments
There are some systems which can use the LLMs themselves to detect writing (basically, if the text matches what the LLM would predict too well, it's probably LLM generated), but they are far from infallible (with both false positives and false negatives). There's also certain tropes and quirks which LLMs tend to over-use which can be fairly obvious tells but they can be suppressed and they do represent how some people actually write.