How old your voice sounds may track brain aging and dementia

Could a recording of you talking say something about how your brain is aging? A study published September 30 in Science Advances tested that idea on 2,928 Spanish speakers in five Latin American countries.

The tool is a 'speech clock.' Researchers trained a machine-learning model, software that learns patterns from examples, to estimate a person's age from their speech. The estimate itself matters less than the gap between how old someone sounds and how old they actually are.

The participants included healthy adults and people with mild cognitive impairment, Alzheimer's disease, and frontotemporal dementia, a form of dementia that often affects behavior and language. The gap differed between these groups. Healthy people had the smallest gaps, and the gaps grew progressively larger in Alzheimer's disease and frontotemporal dementia.

The team then checked whether the voice gap lined up with anything biological. It correlated with phosphorylated tau, a protein tied to Alzheimer's-type brain changes. It also correlated with several epigenetic clocks, which estimate biological age from chemical tags on DNA.

Voices carried traces of people's circumstances, too. Older-sounding speech was associated with less education, financial strain, food insecurity, and poorer access to healthcare. So the clock seems to pick up more than disease, and the study cannot say which way any of these links run.

Speech is cheap and easy to record, which is the practical appeal. If a measure like this proves reliable, it could help flag people who need a closer cognitive check-up. It is not a diagnostic tool. The study captured each person at a single moment, so it cannot show whether an older-sounding voice predicts later decline. Everyone in it spoke Spanish. Before any clinical use, the model needs long-term testing, trials in other languages, and checks in everyday speaking situations rather than research settings alone.