Is there even an incentive to optimize for such signals, though? Em-dashes have been a known indicator of AI-generated text for a good while, and are still extremely prevalent. While someone who doesn't like AI slop and knows and what to look out for will notice and call out obvious AI comments, the unfortunate truth is that the majority of people simply cannot tell, and even among those who can, many don't care.

Obvious AI-generated posts and articles make it to the front page on a daily basis, and I get the impression that neither the average user nor the moderation team see that as a problem at all anymore.