AI tool analyzes voice to estimate aging speed in 3,000 adults

A new artificial‑intelligence tool that reads a person’s voice can estimate how fast they are aging and flag possible dementia risk, researchers reported on Sept. 30 in the journal Science Advances. The system, tested on nearly 3,000 Spanish‑speaking adults, predicts a “speech age” from the way someone talks and compares it with their actual age. When the predicted age is higher than the real one, the individual is more likely to show cognitive problems, dementia or signs of accelerated biological aging.

The work builds on a growing field of “aging clocks” that use biological data to gauge disease risk. Traditional clocks often require brain scans or blood tests, which are costly, invasive and hard to repeat often. “All of these limitations can be overcome with speech,” said Adolfo García, a co‑author and director of the Cognitive Neuroscience Center at the University of San Andrés in Argentina. By analyzing speech, researchers hope to create a cheap, non‑invasive, repeatable measure that could be used in routine health checks.

Scientists drew participants from the ReD‑Lat consortium, a large dementia study covering Argentina, Chile, Colombia, Mexico and Peru. The group ranged from 18 to 88 years old; about half were cognitively healthy while the rest had mild cognitive impairment, Alzheimer’s disease or frontotemporal dementia. Each person completed seven language tasks, such as describing a short video, naming as many animals as possible in a minute, and retelling a story immediately and after a delay. Researchers recorded the sessions, transcribed them and extracted more than 700 speech features, including pauses, speed, pitch, vocabulary richness and emotional tone. A machine‑learning model trained on these features learned to predict chronological age. When the model’s estimate exceeded a participant’s true age, the gap was larger for those with cognitive impairment, and the widest gaps appeared in people with language‑dominant frontotemporal dementia. The study also found that larger speech‑age gaps correlated with poorer social and economic conditions.

The findings are promising but still early. Experts caution that the tool must be tested over time, across different languages and in clinical settings before it can guide individual patient care. If validated, a simple voice recording could become a routine screening method, helping doctors spot early signs of dementia and track the impact of interventions without expensive equipment.