Whether chatting 1:1 or in a chat server I suspect modern day LLM's are not going to be much different. After a while having an overly confident and often incorrect user in the chat is going to get old and people will kick it out with time. LLM's are just a accumulation of knowledge slurped up from the web and much of that knowledge is incorrect misinformation.
Personally I think LLM's + big data were cart before the horse. What should have occurred first was to clean up all the garbage information on the internet before the scraping, regurgitating and effectively plagiarizing the garbage data. Also, get rid of the human mimicry and pandering. It's a bot. Act like a bot. Stop pretending to be my buddy and stop saying every stupid thing I say is a wonderful idea. That just makes me question the money spent on this beast. If a person does that to me I just stop talking to them.
[1] - https://infobot.sourceforge.net/guide-0.43.x.html unsupported, just a piece of history
Give it a couple years..
I'm anxiously awaiting Pew Research's poll results on why people don't ride unicorns to work.