September 15, 2026, 6:12 pm | Read time: 5 minutes

What the Voice Reveals About Aging

The AI model in the study was able to estimate the actual age of participants surprisingly well based solely on a 30-second voice recording. For women, the estimate was on average about four years off the actual age, while for men it was around 4.4 years.

However, the researchers were not only interested in how well the age could be read from the voice. They also found indications that the voice might reflect differences in physical aging: People whose voices sounded unusually old for their actual age differed in several health characteristics from their peers. Thus, the voice could potentially serve as an easily accessible indicator of aging processes. The results were published in the journal “Npj Aging.”¹

How the Study Was Conducted

People of the same age can age physically very differently. Researchers are therefore looking for so-called biomarkers to capture such differences. However, many of the methods developed so far require blood samples, complex laboratory analyses, or medical imaging.

The scientists at the Weizmann Institute of Science wanted to know if a simple voice recording could also provide information about aging. Their hypothesis: Age-related changes in the larynx, respiratory system, muscles, and their control could leave traces in the voice.

For the study, they evaluated data from 6,979 Hebrew-speaking women and men between the ages of 40 and 70. All participants counted aloud from one to 30 for 30 seconds in a standardized environment. An AI voice model analyzed the recordings and determined the so-called “Voice Age,” which is the age estimated based on the voice. The AI model significantly outperformed conventional acoustic measurement methods such as pitch or voice tremor. The researchers then compared this with the actual age and numerous health data of the participants.

Voice Age and Health Are Linked

To examine more closely what an unusually old-sounding voice might mean, the researchers calculated the so-called “Voice Age Acceleration.” Simply put, it indicates whether a person’s voice sounds older or younger than would be expected for their actual age.

There were particularly notable correlations with body composition. A voice that sounded older than the actual age was associated with a larger waist circumference, a higher body mass index, more body fat, and more abdominal fat tissue in both women and men.

The researchers also found correlations with various sleep metrics. These included indications of more frequent breathing pauses or shallow breathing and less favorable oxygen supply during sleep. However, not all of these correlations were statistically significant for both women and men. Some health signals were also gender-specific: For women, an older-sounding voice was associated with lower muscle strength (hand grip strength) and higher blood pressure, while for men, certain liver values were particularly noticeable.

Additionally, the researchers compared the voice analysis with eight other methods of age estimation. The voice performed surprisingly well: Only one method, based on complex metabolic analyses, could predict chronological age better. However, the voice is particularly valuable as a complement: When the researchers combined the voice analysis with other examinations–such as metabolic or sleep analyses–the predictive power increased significantly in almost all cases.

More on the topic

Revolutionary AI Model Researchers Can Predict Risk for 1,000 Diseases Based on Sleep

According to a Study Study Shows Link Between What Fathers Eat and Baby’s Birth Weight

Study Assessment

A strength of the study is the large number of people examined. Additionally, the researchers were able to compare the voice data with extensive medical and biological data.

However, the study has some limitations. Since participants were only examined at a specific point in time and not over a longer period, it is not yet possible to say how the voice changes over the course of aging in individuals. Studies that follow the same people over a longer period would be necessary for this.

Furthermore, only Hebrew-speaking people from Israel were examined. It is therefore unclear whether the method would work equally well with other languages and populations. The voice recordings were also made under uniform study conditions. It is still unclear whether the results would be comparable with different microphones, in other rooms, or with repeated recordings.

Finally, the found correlations do not show cause and effect. The study also cannot answer whether future diseases or the further aging process of an individual can be predicted based on the voice. Long-term studies would also be necessary for this.

According to the authors, the sponsors had no influence on the study. One author reported a paid consulting activity as a potential conflict of interest.

Also interesting: People who grow up multilingual age biologically slower

What Previous Studies Show

Another study from 2020 analyzed 20-second voice recordings from more than 2,000 people with heart failure, who were on average in their late 70s.² These patients were observed for nearly two years. The result: The more noticeable the voice value was, the more frequently the affected individuals died during this period. In the group with the highest values, it was more than half, while in the group with the lowest values, it was just under a quarter. However, several authors worked for the company that developed the analysis technique.

How reliable such approaches are overall was examined in a meta-analysis from 2026. It evaluated 96 studies on voice biomarkers in various diseases.³ The method is considered promising, but most studies were small, and many only assessed participants at a single point in time. Additionally, the accuracy depended heavily on the device used, the environment, and the language in which the recording was made. According to the authors, large long-term studies are needed before the voice can be used in everyday medical practice.