> the cost if something goes wrong would be greater because people have to trust the answers they get from Google.
Google already highlights all sorts of usually-irrelevant garbage in special boxes in the results page. Even a very basic LLM integration would rarely be worse than that.
been using Google Search less and less. i get direct, no SEO-gamed, no ad-ridden, answers from ChatGPT. even the wrong answers point me in the right direction. it looks like a solution that didn't align with an ad-selling machine.
I'm not a fan of Google - I think they have the best search technology but it's horribly corrupted by their ad and data theft focus, but search fills a very different use case than a chatbot. Google can be a question answerer, but that's certainly not what it's best at or what it's most useful for.
Google is an "intelligent portal". In 1998 when the internet wasnt really mapped, search engines actually were used largely for discovery. Now almost everybody knows a double digit number of main go-to sites for things, with a very long tail (say government sites) and google is mostly a portal into these. It doubles as a way to look up random facts, and as long as context isn't required I can see a chatbot doing the same.
Google already highlights all sorts of usually-irrelevant garbage in special boxes in the results page. Even a very basic LLM integration would rarely be worse than that.