> The idea that hallucinations are somehow less likely because you're asking meta-questions about LLM output is completely without basis

Not sure who you're replying to here – this is not a claim I made.