Fact · Voice privacy and security
73percent
In a 2023 PLOS ONE study, listeners correctly spotted speech deepfakes only 73% of the time.
Checked against its source on October 17, 2026
In the source's own words
“Listeners only correctly spotted the deepfakes 73% of the time”
Warning: Humans cannot reliably detect speech deepfakes, Kimberly T. Mai and Lewis Griffin, PLOS ONE, 2023, Abstract.
Open the original (DOI 10.1371/journal.pone.0285333)- What it measures
- Participants listened to real and synthetic speech in English and Mandarin and were asked to identify the deepfakes. The result was the same in both languages.
- What it does not show
- The fakes were made with older algorithms, so newer clones may fool listeners more often. The study tested listening, not a live phone call.