AI 'speech clock' estimates ageing speed from four minutes of speech and rates cognitively impaired voices as older

The study introduces and tests a speech-based ageing clock that estimates a person's age from hundreds of vocal features and defines a 'speech age gap' as the difference between the clock-predicted age and chronological age. Previous ageing clocks relied largely on biological markers, such as 'brain clocks' using neuroimaging signatures and 'epigenetic clocks' using patterns of methyl tags on DNA, whereas a clock based on speech had not yet been developed. The report says the findings were published in Science Advances, with 2,928 Spanish speakers across five countries, including healthy individuals and people with mild cognitive impairment, Alzheimer's disease or other forms of dementia; the model was trained on more than 700 speech characteristics extracted from the recordings.

The speech clock could distinguish healthy individuals from those with some form of cognitive impairment: the speech of people with cognitive issues was categorized as older than expected for their chronological age, while healthy people were generally assessed as having speech matching their chronological age. This links the speech age gap to cognitive status, suggesting the clock could help determine whether a person is growing older faster than expected. The report states that large speech age gaps were strongly associated with cognitive issues such as those arising in dementia, and that the clock could distinguish healthy individuals from those with some form of cognitive impairment; no specific effect sizes or classification accuracy figures are given.

The tool yields predictive value from roughly four minutes of speech recordings and does not rely on expensive or invasive technologies. Compared with brain scans and blood tests, speech recording is easier to deploy in low-resource regions, opening a path to tracking ageing in those settings. Co-author and neuroscientist Agustín Ibáñez says 'we can see a huge predictive value, just with a very simple four minutes of speech recordings'; Jed Meltzer, a cognitive neuroscientist at the University of Toronto not involved in the study, calls it 'a very impressive piece of work' and notes it does not rely on expensive or invasive technologies such as brain scans and blood tests.

The result applies to adults who are native Spanish speakers from Argentina, Chile, Colombia, Mexico and Peru, including healthy individuals and people with mild cognitive impairment, Alzheimer's disease or other forms of dementia; the intended setting is estimating age and computing a speech age gap from about four minutes of speech recordings, so that ageing can be tracked where expensive or invasive tools such as brain scans and blood tests are unavailable. For clinicians and researchers, this offers a speech-based window onto ageing and cognitive status, on which further work can explore how the speech age gap relates to cognitive change.

The report is a summary and does not give the specific effect size linking the speech age gap to cognitive issues, the model's accuracy in separating healthy from cognitively impaired individuals, or which speech features contribute most; the sample is also limited to Spanish speakers in five countries, so performance in other languages and populations remains an open question. These missing details affect how strongly the conclusion can be judged and are worth watching for in the original paper.