Scientists have introduced an artificial‑intelligence tool that estimates a person’s biological age by analyzing characteristics of their voice. The so‑called “speech clock” evaluates hundreds of vocal parameters, such as pitch, speaking speed and vocabulary range, to produce a predicted age that can be compared with the individual’s chronological age.
How the speech clock works
The system draws on more than 700 acoustic and linguistic features extracted from short recordings—typically four minutes of speech per participant. By feeding these data into machine‑learning algorithms, the researchers trained a model that can output an age estimate based solely on vocal output. The difference between this predicted age and the person’s actual age is termed the “speech age gap.”
In the study, large speech age gaps were strongly linked to cognitive problems, including those associated with dementia. Participants whose speech was assessed as older than their calendar age tended to exhibit signs of mild cognitive impairment, Alzheimer’s disease or other dementias, whereas healthy volunteers generally showed speech ages that matched their true ages.
Potential applications and advantages
According to neuroscientist Agustín Ibáñez of Adolfo Ibáñez University, the method offers “huge predictive value” despite requiring only a brief speech sample. The approach could be especially valuable in low‑resource settings, where expensive brain imaging or blood‑based tests are not readily available. Cognitive neuroscientist Jed Meltzer, who was not involved in the work, described the technique as “a very impressive piece of work” that may enable large‑scale monitoring of ageing trajectories without invasive procedures.
Existing biological clocks rely on markers such as neuroimaging signatures (“brain clocks”) or DNA methylation patterns (“epigenetic clocks”). The speech clock adds a new, non‑invasive dimension to this field, capitalising on the fact that speaking engages extensive brain networks.
Study design and validation
The researchers recorded 2,928 Spanish‑speaking volunteers from Argentina, Chile, Colombia, Mexico and Peru. The cohort included both cognitively healthy individuals and patients diagnosed with mild cognitive impairment, Alzheimer’s disease or other forms of dementia. Participants performed a series of speech tasks while their voices were captured for analysis.
Machine‑learning models were then trained on the extracted acoustic features to predict each participant’s age. The resulting speech clock was able to differentiate between healthy and cognitively impaired groups, consistently assigning older‑appearing speech ages to those with cognitive deficits. The findings were published in the journal Science Advances.
While further research is needed to confirm the tool’s utility across languages and clinical settings, the initial results suggest that voice‑based age estimation could become a low‑cost, scalable method for early detection of accelerated ageing and neurodegenerative risk.
Steve Lopez is a Senior Editorial Columnist and Health & Public Policy reporter for News Raise. Steve focuses on healthcare advancements, medical technologies, and public health policies.
