Runway's "Turing Reel" Study Finds 90% of People Cannot Distinguish AI-Generated Video from Real Footage
Published 2026-01-22Ingested 2026-01-31Foundation ModelsMedium
Summary
Runway published "The Turing Reel" research on January 22, 2026, a study evaluating whether people can distinguish AI-generated video from authentic footage. The study recruited 1,043 participants, each of whom viewed 20 five-second videos (10 real, 10 generated by Runway's Gen-4.5 model) in randomized order across five content categories: faces, full-body human motion, animals, nature scenes, and urban environments. Key findings: Overall detection accuracy was 57.1%, only marginally above the
Alignment: Reinforces current position
Related Positions: ai-governance-and-risk.md
runwaygen-4.5ai-videodeepfakesdetectionturing-testsynthetic-mediacontent-authenticationc2pagovernancemisinformationresearch