What’s worked well for us there is the fact that we’ve (forever!) had “Be Concise” as one of our top-line community standards. LLM output is frequently the exact opposite of concise.
I’ve had similar irate conversations, but by making it clear why the LLM output is problematic has helped significantly. Somehow the critique on the means of text generation is taken very personally, but the critique on the content is not. I suppose that’s not terribly surprising — they didn’t write the content!