LLMs are best used as a thought partner, not as an oracle.

They will often be wrong; you should assume anything they tell you is likely to be approximately right, but perhaps wrong in the details.

But if you use them as a generative thought partner, they can be very generative!