Large models in general. A semantic query for "fake articles", without examples, is a wildcard search.

A commercial RAG solution would use Query Expansion (QE) and examples to find nearest neighbors.