The synthetic media campaign was identified through a combination of platform integrity reporting and independent open-source analysis. An initial signal came from a civil society organization in Brazil that had tracked a network of social media accounts amplifying narratives about a candidate in the lead-up to a federal election. SCRYER analysts were engaged to assess attribution and campaign architecture. Persona graph analysis of 284 identified accounts revealed three distinct operational tiers: a seed layer of 18 accounts with multi-year posting histories and plausible personal narratives, a mid-layer of 94 accounts created between one and six months before the election whose content was exclusively political, and an amplification layer of 172 accounts with no original content that served purely as signal boosters. The seed accounts showed evidence of AI-assisted persona construction: profile images generated by a diffusion model, posting histories with subtle chronological inconsistencies, and biography fields that combined authentic cultural references with statistically unusual occupational combinations.

Platform amplification mechanics relied on coordinated cross-posting timed to achieve algorithmic promotion during high-engagement periods. The campaign operators had mapped the recommendation thresholds of five major platforms and calibrated the amplification layer's activity to trigger trending signals without reaching the engagement volumes that would activate platform integrity response teams. Synthetic video content depicting a candidate making statements they had not made was distributed exclusively through private messaging groups rather than public channels, limiting platform detection while maximizing reach within targeted demographic segments. Reach estimation of 8.6 million unique users is a conservative lower bound based on confirmed account follower counts and platform sharing data provided under legal process. The actual reach, including resharing by authentic users who encountered the content organically, is assessed to be significantly higher.

Detection signal recommendations for platform integrity teams are organized around three tiers of reliability. Tier-one signals, those with low false-positive rates and high automation potential, include diffusion-model image fingerprinting at account registration, chronological consistency scoring for posting histories, and cross-account coordination detection via temporal clustering of identical or near-identical content. Tier-two signals, requiring human analyst review, include persona narrative coherence assessment and behavioral timing analysis against platform-specific algorithmic thresholds. Tier-three signals, requiring external coordination, include cross-platform account correlation and civil society network reporting integration. The 30-day gap between campaign detection and platform enforcement action documented in this case is consistent with the structural delay in most major platforms' policy enforcement pipelines, and represents a critical window during which synthetic media campaigns can achieve their primary amplification objectives.

Campaign operators had mapped the recommendation thresholds of five platforms and calibrated the amplification layer to trigger trending signals without activating platform integrity response teams.