[mplewis]

The LLM is not a human and cannot distinguish between spam and high quality content.

[serjester]

It seems pretty trivial to fine tune a model to do this - this is really playing to LLM's strengths.