if you are dumb and looking just for confirmation of your bias, superstition, you will sooner or later find something to confirm that. whether it is chatbot, google, tv, or paper news. the tool can’t fix dumb person using it.

I think you may be oversimplifying what's happening here. For the sake of argument, let's say LLMs are just another tool - a pretty common (and I believe false) response. But even if true, the problem with saying "it's just another incremental difference" is that incremental differences do add up. For AI, there is a material difference between the Google situation and a chatbot, even if incremental.

AI specifically tailors its response to you in a way that is not just syncophantic. It "logic-washes" substantial uncertainty and credibility issues. When you click on some crazy person's blog on Google, you may in fact conclude that lizard people are not running the world as they claim because the grammar is bad, the layout is schizophrenic, logical leaps abound. But when you talk to a chatbot, you can ask them to explain why lizard people are real, and they may explain in very logical-sounding terms, in language you only have seen previously from authoritative sources in your life, in an organized layout, why lizard people are real.

People have confirmation bias, but to take your "the knife is not responsible if you kill someone with it": a nuclear bomb also isn't responsible if it flattens a city. But that doesn't mean there isn't a material difference in the inherent dangers of its existence, regardless of who uses it.