Overview
- Researchers published the study on Sept. 30, 2026 showing a speech-based model trained to predict a person’s chronological age from short voice recordings.
- The model was trained on recordings from 2,928 Spanish-speaking adults across Argentina, Chile, Colombia, Mexico and Peru and used more than 700 extracted vocal features.
- Investigators calculated a “speech age gap” — the difference between speech-predicted age and actual age — and found larger gaps distinguished people with mild cognitive impairment or dementia from healthy participants.
- Follow-up analyses linked larger speech age gaps to higher dementia likelihood on brain scans and to higher plasma p-tau217 in people diagnosed with Alzheimer’s, but those biomarker findings are preliminary.
- Authors and outside experts say the clock could allow low-cost, scalable screening with only a few minutes of speech, yet they stress the need for replication across languages, control for education and recording conditions, and discussion of privacy and ethical issues before clinical use.