I find it's great at summarizing a field, or giving a very generalized "look over there" kind of answer to top-level questions about scientific inquiry.
But I've found they (at least Google Gemini) can mess up sometimes... Like, citing something that actually shows the opposite of what they claim.
I think they're very incentivized to give "an answer", when there are important situations where the correct answer is "we don't know" or "your question doesn't make sense," etc.