Once they became a serious product pushed by major tech companies, people aren't unreasonable in expecting it to be right most of the time. Especially if it's going to go alongside or replace traditional search.
Tech executives are betting on the hype first and hoping their talent will make it into a usable product.
Even it the chatbot improves, I imagine many folks are waiting for the point where they have some direct control of its training / fine tuning that is easy and accessible. You can correct it, but why waste time doing that if it won’t remember next time? Issues like, e.g. if you have NDAs, are you breaking them by disclosing information with a chatbot? There’s little clarity in regards to those issues, and you can’t just take someone’s word for it. Ideally you’d want control of the chatbot’s learning process and how it handles all kinds of (sensitive?) data.
Tech executives are betting on the hype first and hoping their talent will make it into a usable product.