Voice Age Test: Can Your Voice Reveal Your Biological Age?

What is a voice age test?

A voice age test estimates how old a person is from a recording of their voice, and researchers compare that estimate with the person's real age to see whether the voice sounds older or younger than expected ^[1][2]. The gap between the two numbers is the part scientists are interested in.

In research, the estimate comes from a computer model trained on thousands of recordings from people whose ages are known ^[1][2]. The model learns which features of the voice tend to change with age, such as pitch, timing, pauses and, in some studies, the words people choose ^[2][3].

The idea follows other 'ageing clocks', which estimate biological age from DNA markers, brain scans or blood tests ^[2][3]. If someone's clock age is higher than their age in years, the clock suggests their body may be ageing faster than average; if it is lower, it suggests the opposite.

Two large studies published in September 2026 tested this idea with voice and speech: one in Israel and one in Latin America ^[1][2]. They used different recordings, languages and methods, so their results are not directly comparable, but together they show where research on voice and age stands in November 2026.

What did the Weizmann Institute voice age study find?

Researchers at the Weizmann Institute of Science found that a 30-second voice recording contained a reproducible age-related signal in 6979 Hebrew-speaking Israeli adults aged 40-70 ^[1]. The study, by Krongauz, Marmor, Segal and colleagues, was published in npj Aging on 11 September 2026 ^[1].

The team trained separate models for women and men. The models estimated chronological age with an average error of 3.95 years in women and 4.41 years in men, and explained 53.9% and about 44% of the variation in age respectively ^[1]. The researchers call the result 'Voice Age' ^[1].

Voice Age was the second-best single source of information among nine age models the team compared, which also used molecular, imaging, physiological and lifestyle data ^[1]. It only partly overlapped with those other clocks, and adding it to them improved age prediction in nearly all comparisons ^[1]. Combined with blood metabolite data, the models explained 65.1% of the variation in age in women and 52.2% in men ^[1].

People whose voices sounded older than their age tended to have more body fat, more sleep-disordered breathing, lower night-time oxygen levels and differences in liver imaging measures ^[1]. Some links, such as grip strength, differed between women and men ^[1]. The paper notes that one author, Eran Segal, is a paid consultant of Pheno.AI Ltd ^[1].

What did the Latin American speech clock study find?

A team whose first author is Hernán Hernández of the Latin American Brain Health Institute (BrainLat) in Chile built a 'speech clock' from 2928 native Spanish speakers in Argentina, Chile, Colombia, Mexico and Peru ^[2][5]. It was published in Science Advances on 30 September 2026 ^[2][4].

About half of the participants were cognitively healthy: 1504 healthy adults, 24 people with mild cognitive impairment, 1068 with Alzheimer's disease, 255 with non-language frontotemporal dementia and 77 with language-dominant frontotemporal dementia ^[2]. Overall, 67% were women and the average age was 65.23 years ^[2].

Each person completed seven standard speaking tasks, such as viewing and describing a short animated film and retelling a story ^[2]. The researchers extracted more than 700 features from each recording and its transcript, covering timing, pitch, word choice, emotional content and how much people said ^[2]. The model estimated age with an average error of 9.04 years ^[2].

The 'speech age gap' was smallest in healthy participants and larger in Alzheimer's disease, and larger still in frontotemporal dementia ^[2][3]. In people with Alzheimer's disease, a larger gap went with higher blood levels of p-tau217, a protein linked to the disease ^[2][3]. Pitch and emotional content were most linked to an older-sounding profile, while timing and the number of words were linked to a younger one ^[3].

How accurate is a voice age estimate?

On average, the Israeli voice model missed a person's real age by about 4 years and the Latin American speech clock by about 9 years ^[1][2]. These are averages across thousands of people; for some individuals the error is much larger.

The two numbers should not be compared head to head. The Israeli study included only adults aged 40-70, a narrower range, while the Latin American study covered ages 18 to 88 and included many people with dementia ^[1][2]. The recordings were also different: a 30-second recording in one study, seven structured speaking tasks in the other ^[1][2].

For research, the error is less important than whether the gap between voice age and real age says something about health. Both studies found that it does at group level: the gap was linked to body measures in Israel and to dementia and brain markers in Latin America ^[1][2].

An average error of 4 to 9 years also shows why a single test result means little for one person. A 50-year-old whose voice is estimated at 55 may simply fall within the normal error of the model rather than be ageing faster.

What can't a voice age test tell you?

It cannot tell you how fast you personally are ageing or whether you will develop a disease, because both studies looked at people at a single point in time ^[1][2]. The Science Advances authors write that their design does not establish whether speech age gaps reflect faster ageing over time ^[2].

The NIH, which helped fund the speech clock research, said long-term studies are needed to find out how well speech-based clocks can predict ageing-related problems such as dementia ^[3]. Medscape's report described the speech clock as a candidate marker for research on ageing and dementia, not a clinical test ^[5].

There are also technical limits. The Latin American authors say that differences in microphones, recording conditions and speaking tasks can add noise, that speech tools may carry dialect or cultural bias, and that their results apply to Spanish-speaking cohorts and still need testing in other languages ^[2]. The Israeli study used Hebrew speakers aged 40-70 ^[1].

The Latin American authors also note that sex, other health conditions and mood may influence speech age gaps ^[2]. This guide does not give medical advice: anyone worried about changes in their voice, memory or health should talk to a qualified health professional.

Are online 'how old do I sound' tests the same thing?

No. Free online voice age tests are separate consumer tools; they are not the models from the npj Aging or Science Advances studies ^[1][2][6]. One such site asks users to read a short passage aloud and then gives an age estimate with an explanation ^[6].

That site itself says the result is a best guess, not a medical or biometric measurement, and describes it as a fun reflection of how your voice comes across ^[6].

The research models were trained and checked on thousands of people with known ages and, in some cases, linked medical data ^[1][2]. The research models are described in peer-reviewed papers with their error rates ^[1][2]; the consumer site checked here describes its result only as a best guess ^[6], so its numbers cannot be compared with the studies.

If you try one for fun, treat the number as entertainment. A high or low score is not evidence of good or poor health.

Why are scientists interested in voice as an ageing marker?

Because a voice recording is cheap, quick and does not need a needle or a scanner, it could make ageing research possible in places where brain scans and DNA tests are hard to get ^[2][3][5]. The NIH summary says the approach could become a less expensive and easier way to measure how ageing affects the body and brain ^[3].

The Latin American team stresses this point: existing clocks rely on MRI scans, blood samples or tests given by trained staff, while speech can be recorded and transcribed automatically and remotely, according to Medscape's report on the study ^[5]. The authors describe speech age gaps as a possible low-cost marker for research in under-represented regions ^[2].

The Israeli team argues that voice adds information that other clocks miss, since Voice Age only partly matched molecular and imaging clocks and improved them when combined ^[1]. Its strongest links were with body fat, breathing during sleep and liver measures ^[1].

Speech results also matched other clocks in the Latin American study: speech age gaps were associated with brain-age gaps from MRI and with epigenetic age estimates from DNA in blood ^[2][5].

What should you check before trying a voice age test?

Check who made the test, what it measures and what happens to your recording ^[6].

1. Look for a study: research results come from peer-reviewed papers such as those in npj Aging and Science Advances ^[1][2].
2. Read the disclaimer: a consumer tool may say plainly that it is not a medical measurement ^[6].
3. Protect your voice data: a voice recording can identify you, so check whether it is stored or shared.
4. Note the recording conditions: differences in recording conditions and microphones can add noise to speech measurements ^[2].
5. Do not rely on one result: even the best research model missed real age by about 4 years on average ^[1].
6. Talk to a professional about health worries: no voice test replaces a medical assessment ^[3].

As of November 2026, neither study presents a test for personal use ^[1][2].