Study Shows AI Beats Humans On Deepfake Images

Psychologist Natalie Ebner and colleagues report January 7 in Cognitive Research that AI systems outperform humans at detecting deepfake images, while humans outperform algorithms on short deepfake videos. In tests with about 2,200 participants and two classifiers on 200 faces, algorithms reached about 97% and 79% accuracy; in a 70-video test with ~1,900 participants humans averaged 63% versus algorithmic chance.
Key Points
- 1Show AI detects deepfake images with roughly 79–97% accuracy versus human chance-level performance
- 2Reveal humans outperform AI on short deepfake videos, averaging 63% versus algorithmic chance
- 3Recommend combining human intuition with AI detectors to strengthen deepfake detection workflows
Scoring Rationale
Peer-reviewed evidence shows mixed human–AI performance, but findings are limited to specific datasets and short videos.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems

