We cross-check creator content, audience and behavior. Signals include sudden follower spikes, inconsistent engagement vs. follower size, bot-like comment patterns, geographic mismatches, and improbable view distributions. Semantic AI and rolling L30D metrics flag anomalies. Results combine fraud scores with engagement and audience verifications.