As good as a coin toss: Human detection of AI-generated images, videos, audio, and audiovisual stimuli
Source record
cooke-2024-coin-toss· 2025 · Communications of the ACM
Evidence tier: B (peer-reviewed-journal) · Verification: verified-abstract (compiler confidence: high)
Bottom line: Mean detection was close to 50% chance; accuracy dropped for any synthetic content, foreign-language content, single-modality stimuli and synthetic images of faces; prior knowledge of synthetic media did not help; older participants did worse.
Citation
Di Cooke, Abigail Edwards, Sophia Barkoff, Kathryn Kelly (2025). As good as a coin toss: Human detection of AI-generated images, videos, audio, and audiovisual stimuli. Communications of the ACM, 68, 100-109. https://doi.org/10.1145/3729417
Verification entry (identifiers, resolvers, claims with source locations): /data/citations.json#cooke-2024-coin-toss
Study design
| Field | Value |
|---|---|
| Design | Online US survey; mixed-modality stimuli (images, audio, video, audiovisual) |
| Modality | image-face, video, audio, audiovisual |
| Stimuli | Mixed 2023-era generators |
| Participants (N) | 1276 |
| N detail | preprint first posted March 2024 (arXiv 2403.16760) |
| Topics | human-detection, modality |
Results
| Metric | Value | Note |
|---|---|---|
| Mean detection | close to chance (50%) | abstract gives no exact figure |
| Lower accuracy when | stimulus contains synthetic content, foreign language, single modality, or a synthetic human face | |
| Mixed-authenticity audiovisual | harder | |
| Prior knowledge of synthetic media | no significant effect | |
| Age | older participants performed worse |
Limitations
- US sample only
Cited in
- MR-0001: Human Detection of Deepfakes: Accuracy, Bias, Confidence, and Training
- MR-0003: Voice Clones and Synthetic Speech: How Well Do People Detect Them?
Verification
| Field | Value |
|---|---|
| Status | verified-abstract |
| Compiler confidence | high |
| Checked on | 2026-10-07 |
| Via | arXiv 2403.16760 abstract page (journal ref CACM 68, 2025) |
Values on this page have not all been checked against the full text. Confirm them against the primary source before high-stakes use.