LLM generated text is fascinatingc because there's two groups managing it.
1) I am going to call the SpamAssassin side, now I know that you've likely not has email that wasn't Office365 or Gmail for a decade or more, and those are Category Two email services, but we scrappy underdogs filtering email with SpamAssassin have an effect where LLM generated text looks like garbage, and so if you generate your words or images with a GPT engine, it's gotta be scored as garbage
2) the Google/Facebook/Microsoft stance that you should allow all LLM text and diffusion images! They're great! It's the AI revolution!

The problem with Category 2 is that it rapidly makes all your stuff garbage (people only see spam), because if people can generate stuff no one wants to see, but gets past the filter, they will. Google will literally pay them to do that.

If you block that stuff, people who claim there are ethical uses for LLMs (a lie) will be caught, and those generating human supervised text carefully will be blocked.

The goals of people-who-make-money-on-clicks (read: everyone) are only served if:
We only use category 1 filters
The people making money on clicks create things that pass category 1 filters

And the whole "what is happening to facebook" is just the culmination of this obvious point