Can people really tell the difference between a real human voice and one generated by AI?
According to a new study, your conscious mind might struggle, but your brain is already picking up the clues. Researchers discovered that listeners are generally poor at distinguishing “deepfake” speech from genuine human voices—even after brief training. However, brain activity revealed something surprising: with minimal training, neural responses to AI-generated speech and human speech became noticeably more distinct. This indicates that our auditory systems may be adapting to AI voices faster than our conscious decision-making.

Key Findings

• Conscious Limitation: Participants had difficulty reliably identifying AI versus human voices, and short training sessions only slightly improved their accuracy.
• Neural Sensitivity: Brain measurements showed a clear increase in differentiation after training. The brain began responding differently to AI speech compared to human speech, even when participants couldn’t explain the difference.
• Hidden Acoustic Cues: The auditory system appears sensitive to subtle micro-acoustic signals—small irregularities in rhythm or tone typical of AI voices—that people don’t consciously notice.
• Human Adaptation: The results suggest humans are currently adapting to AI-generated content. Our brains detect useful cues, but we haven’t yet learned to consciously use them.
• Fraud Prevention Potential: These findings could help guide the creation of better methods and training to help people recognize deepfake scams and voice-cloning fraud.

To the study