Most scam-detection tools flag keywords. This one thinks like an investigator: a custom Naive Bayes classifier filters out defensive PR spin before scoring, weighted trust by source authority (govt records > social posts), and enforces relevance so results actually match the entity—not just its name. Wrapped in a Cyber-Noir UI with real auth and OSINT via SerpApi, it's built as production software, not a hackathon script.