Comment by idiotsecant

The product need here is an LLM 'answering machine' that accepts emails, determines if the email is spam or a valid email or something in between. If It's in between it could say something like 'Hello, I am [user]s email filtering service. I think your email might be automatically generated, can you please tell me why [user] would need to see this? and make a judgement based on the response.

No spammer will manually reply to that, some AI spambots might, but it should be apparent to the LLM that's what is happening.