To be fair I treat AI in the same way I treat other human beings; if claims are made then they better be backed up by a reputable source.

ChatGPT is very good at giving me a link to a source. Friends making outrageous claims, not so much.

That seems like a reasonable approach, if you're at least skimming the source afterwards to verify.

It'd be good if most people did that as a bare minimum, though it doesn't seem to be how it's often done currently.