The Scale of the Deepfake Problem in 2026

Non-consensual deepfake content has grown at a compound annual rate exceeding 900% since 2019. The majority of deepfake content, estimated at over 96% by volume, is non-consensual intimate imagery targeting private individuals, not public figures. Women account for approximately 90% of identified victims.

Platform Distribution

Deepfake content is not confined to a single platform. The top distribution channels include dedicated deepfake sites, general pornography aggregators, Telegram channels, Reddit communities, and Discord servers. Content spreads through re-upload networks that make single-platform removal insufficient, a piece of content removed from one site typically reappears on 3 to 7 others within 48 hours.

Detection Gap

Standard reverse image search tools find fewer than 4% of non-consensual deepfakes in circulation. This is because most deepfakes are re-encoded, cropped, or filtered before distribution, breaking the pixel-level hash matching that standard tools rely on. Biometric face-embedding matching, the approach used by ScanErase, detects re-encoded and altered copies that reverse image search misses.

What the Data Means for Victims

Most victims are unaware content exists until it is discovered by a third party. Early detection dramatically improves removal outcomes, content reported shortly after first publication is significantly easier to contain than content that has circulated for weeks. Content that has circulated for more than 30 days has a lower same-week removal rate due to re-upload proliferation across multiple platforms.