We run semantic analysis on video titles, descriptions and transcripts, then cross-check audience signals - gender, geography, language, reachability - against fraud and engagement metrics (ER%, recent views, growth, posting cadence). For YouTube couple influencers we flag anomalies, normalize views and surface true reach in one profile sheet.