In many ways it's worse, because with social media you know -- or think you know -- the source, and you use that to judge accordingly how open you are to what you're reading or how much you trust its veracity, or what biases it might have. Whereas with LLMs most people inherently trust because the LLM is supposedly objective and unbiased.

Id argue its already even greater. The amount of botting and astroturfing happening now means that a good chunk of the content people are consuming on those social media sites is generated by LLMs. Social media is the vector, but its the content thats the virus and OpenAI et al control what that virus contains

That's a really good point. Even if social media is and remains the medium over chat assistant apps (likely imo), all the content everywhere will be AI slop.