Fact · Voice privacy and security
The PLOS ONE study found that giving listeners examples of speech deepfakes improves their results only slightly.
Checked against its source on October 17, 2026
In the source's own words
“Increasing listener awareness by providing examples of speech deepfakes only improves results slightly.”
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
- The same experiment, with a training step in which listeners heard examples of speech deepfakes before the test.
- What it does not show
- A small gain from one training session. It does not show that training can never help.