Problem at scale. Doesn't matter if someone is consciously able to identify individual bot accounts or comments. There can still be a strong general feeling that something is very wrong. Leading to more and more frustration and unhappiness.

"Popular" reddit posts and subreddits are a good example of this.

It might be hard to recognize an individual user or post as AI, but it's not hard to recognize the negative effect in aggregate