i actually think this probably has a lot to do with tone and hype. wikipedia has always been a fantastic resource, but it doesn't advertise, and it's standards require sources to make a claim AND a polite, neutral tone. they don't ALWAYS succeed, but they work really really hard to avoid as much bias as possible. wikipedia isn't here to convince you of anything or manipulate you, it's just stating facts for you to do with what you will.
meanwhile, ai has been absolutely hyped beyond belief. we've been inundated with it literally everywhere; i'm currently trying to buy a new stove and most of the fucking things have an ai assistant built in. and to here the tech companies talk about it, you'd think the ai is some sort of hyper-intelligent god they created. they act like it's fucking deep thought from hitchhikers guide to the galaxy. they will say literally anything to get people to Buy Product.
and add to that, ai isn't even designed to be HELPFUL, it is designed to be PROFITABLE. to keep people coming back. so they don't care if it tells the truth, they just care if you are CONVINCED that it's smart and cool. so it's designed to sound confident and charismatic and perfectly tailored to what you, personally, find appealing at all times. they couldn't have made a better disinformation machine if they set out to do so.
this is sorta like asking "oh, so you don't believe the dry research papers, but you DO believe the con man who has already tricked hundreds of people?" like, yeah, we'd all like to believe that we'd pick the scientific evidence, but it's very very easy to fall for a con. now imagine if that con had access to all your data and was ever changing to reflect your personal tastes and history and current emotional state. you can see why we have a problem.
i think it might also have something to do with the fact that wikipedia has an overt human element to it. a lot of the emphasis on its unreliability has come from the part where anyone (with an account) can edit an article, and that has indeed caused them some grief in the past.
more importantly though, i think the attitude around llm platforms is āthe machine is very smart, it has a lot of informationā, and the perceived impartiality of a computer program makes people think itās infallible, or at least mostly reliable.
people think the machine can do it better than humans because humans make mistakes, but they arenāt realizing that itās not an answers machine, itās a guessing machine, meanwhile the human sources have trained themselves for over two decades now to be diligent enough to not let any mistakes into their answers
