Having reluctantly used both, Bing's Copilot seems a lot more grounded on current search results below it versus Google's Gemini seems a lot more likely to conduct its own searches from a different query than what was asked, so also a lot more likely to hallucinate things or to provide answers that seem way different from the rest of the search page.
In terms of "best on the market" for AI search, I know that I am much more likely to trust the one that seems more like a direct summary of the stuff the search engine is traditionally responding with (and presumably has been well tuned in the last several decades) versus the one more likely to make stuff up or to leave the realm of what you are actually asking for some other search it thinks is better for you.
Though admittedly that's a very personal judgment call; some people want the assistant to search for "what they really mean" rather than "what they asked for". It's also a lot of gut vibes from how these AIs write about their research and some of that can be hallucinations and lies and "prompt optimization" as much or more than any sort of "best on the market" criteria.
In terms of "best on the market" for AI search, I know that I am much more likely to trust the one that seems more like a direct summary of the stuff the search engine is traditionally responding with (and presumably has been well tuned in the last several decades) versus the one more likely to make stuff up or to leave the realm of what you are actually asking for some other search it thinks is better for you.
Though admittedly that's a very personal judgment call; some people want the assistant to search for "what they really mean" rather than "what they asked for". It's also a lot of gut vibes from how these AIs write about their research and some of that can be hallucinations and lies and "prompt optimization" as much or more than any sort of "best on the market" criteria.