A speech clock reads biological ageing from 2,928 voices

A team led by Agustin Ibanez at the Global Brain Health Institute and Trinity College Dublin reported on 30 September 2026, in Science Advances, a machine-learning speech clock that estimates age from the way a person talks. The models were trained on acoustic and linguistic features of speech from 2,928 Spanish-speaking people in Argentina, Chile, Colombia, Mexico and Peru, including healthy adults and people with dementia.

The measure of interest is the speech age gap: the difference between the age the voice predicts and the person’s actual age. Healthy participants showed the smallest gaps, and the gaps grew across Alzheimer’s disease and forms of frontotemporal dementia.

The gap tracked independent markers of ageing. It correlated with MRI-measured brain atrophy and cortical thinning, changes in functional connectivity, three DNA-methylation clocks, plasma p-tau217 in Alzheimer’s disease, executive function and working memory, daily functional independence, and lifelong social adversity.

The authors describe the speech clock as an investigational research framework rather than a diagnostic test. They say prospective longitudinal studies are needed to establish whether it predicts future cognitive decline.