Em dashes (—) used to be a rough litmus test for detecting AI generated content. Even though there are many false positives to this classifier, it is in general a good feature in predicting confidence for AI slop text.

Since then, em dashes have evolved. Here's the new AI slop predictor:

"That's not just X. That's Y."

It comes from the latest OpenAI models. I have read this over and over on linkedin (aka cesspool of AI slop text).

That's not just a predictor of AI slop. That's a predictor for what you should or shouldn't give your attention to.