[lucasluitjes]

Ironically if you wanted to build that accurately and quickly, you would probably end up having an LLM classify content as being LLM-related or not. Keyword-based filtering would have many false positives, and training a model takes more time to build.