I can deal with an ai being wrong because an ai is not malicious. It may be programmed a way but it’s not directly trying to screw me.

You're right, but it's working with data from people who might be, and because it cooks it all down you can't be sure of the origins of that data. There's an entire industry popping up now to "lean" AI answers in "approved" directions. It's SEO optimization, for AI answers.

I treat "AI" like an overhyped, incredibly-energy-inefficient search engine. I check every citation it makes, every time. Sometimes this is better than a straight search, sometimes it's not. But the main thing is that EVERY result is treated as untrustworthy by default. If a boring search engine result persistently sent me to the same site written by a human, and I found the results worthwhile, that site (and the responsible human) would build trust over time and I'd do less checking. But with AI you're constantly dealing with the "data slurry", so no trust is (or ever should be) built. You're starting from scratch, every time.