Fact · Voice privacy and security
27percent
Because listeners spotted 73% of the speech deepfakes in the study, about 27% went unnoticed.
Checked against its source on October 17, 2026
- What it measures
- Calculated from the study's detection rate of 73 percent, which is the share of deepfake clips listeners identified correctly. The remainder were missed.
- How it is calculated
- 100 - 73 = 27