Sometimes I think it's worse even than that. As you say, providing a supporting citation gives a gloss of authority and acceptance, whether in an academic paper

Sometimes I think it's worse even than that. As you say, providing a supporting citation gives a gloss of authority and acceptance, whether in an academic paper or on a chat forum. The reader should ideally appraise the cited source and decide if they agree. I doubt this happens in even 1% of cases. Ideally I'd like to see citations that both support and disagree with the point being made. Then I could take a more balanced overview and decide my level of trust in the evidence. However this would lead to a systematic review for every sentence written/read. To my mind this is a major problem in the peer-review scientific process, and I don't know of a decent solution. This is where I'd like to see a really potent AI: systematic reviews.