The way you speak may reveal more about your aging brain than you might expect. Researchers trained an AI model to listen to people talking and estimate their age, but the model’s mistakes turned out to be especially interesting.
When it guessed someone was older than they actually were, other measures of aging and cognitive health often pointed in the same direction.
The study, published September 30 in Science Advances, used recordings from people with and without dementia diagnoses across Latin America. Researchers call the difference between the model’s estimate and a person’s chronological age the “speech age gap.” That gap was associated with several measures of aging that ordinarily require brain scans, blood tests, and cognitive assessments.
It is an intriguing idea, especially because recording speech is so simple. Still, this is not a voice-based dementia test, and the researchers make that distinction clear. What the study found is subtler: everyday speech may contain clues about biological and cognitive aging that scientists are only beginning to understand.
What The Algorithm Actually Heard
The model weighed the words people chose along with the sound of their voices.
The researchers recorded 2,928 Spanish-speaking participants in Argentina, Chile, Colombia, Mexico, and Peru while they completed three structured tasks. They described a short video, named as many animals as possible in one minute, and retold a story immediately and again after a short delay.
From those recordings, the model extracted hundreds of features. Some involved how people sounded, including speech rate, pauses, pitch, and emotional tone. Others focused on what they said, such as vocabulary, word precision, how much they spoke, and how they organized their language.
The model was trained to estimate chronological age. If its estimate exceeded the person’s real age, it produced a positive speech age gap. A negative gap meant the speech profile appeared younger.
Participants included healthy controls as well as people with mild cognitive impairment, Alzheimer’s disease, and two forms of frontotemporal dementia, including one dominated by language problems.
Where The Aging-Sounding Gap Showed Up
Brain scans, DNA-based clocks, and memory tests all pointed in the same direction.
Among people with Alzheimer’s and both frontotemporal dementia groups, larger speech age gaps were associated with older brain age measured through structural and functional imaging. The gaps also tracked DNA-based aging clocks in healthy controls and people with Alzheimer’s.
Cognitive testing added another layer. Older-looking speech was associated with poorer global cognition, executive function, daily functioning, and several types of memory, including tests that did not depend on language.
For people with Alzheimer’s, larger gaps also went with higher plasma p-tau217, a blood biomarker Trinity College Dublin describes as important for Alzheimer’s pathology.
The groups differed, too. Healthy controls had smaller gaps than participants in the patient groups. Within the Alzheimer’s and frontotemporal dementia groups, the gaps were smallest in Alzheimer’s, larger in the non-language form of frontotemporal dementia and largest in the language-dominant form.
Social Circumstances Showed Up In The Speech Too
The age gaps also tracked the social conditions people had lived with.
Education, finances, food insecurity, health care access and early-life experiences were among the factors associated with older-looking speech profiles in healthy controls and people with Alzheimer’s.
That finding broadens the picture beyond the brain alone. Senior author Agustin Ibanez, a professor in brain health at Trinity College Dublin, says researchers are learning just how much information speech can carry.
“It captures both the passage of chronological time and signals coming from cognition, the brain, systemic biology, and even our accumulated social environment.”
The setting is important as well. The participants came from five Latin American countries, a region Trinity describes as historically underrepresented in dementia research. That gives the study a broader cultural context than research built entirely around populations that have traditionally dominated aging studies.
Save this article
Enter your email address and we'll send it straight to your inbox.
Why It Is Not A Dementia Test
Patterns across measurements do not predict outcomes.
The study was primarily cross-sectional, meaning participants were assessed at one point in time. That allows researchers to see which measurements tend to move together, but it cannot show whether a large speech age gap predicts who will develop dementia or how quickly someone’s brain is changing.
It also cannot prove that difficult social circumstances cause a person’s speech to appear older.
“This study was a one-time snapshot and needs validation in other languages and real-world settings before clinical use,” said James Alrassi to Everyday Health.
There are practical unknowns as well. Trinity’s announcement does not say how accurate the individual age estimates were or how large the associations were. It also mentions no app, approved service, or launch date.
Related: 10 Simple Habits to Recharge Your Mind and Body
Should You Worry About How Your Voice Sounds?
Most voice changes with age are ordinary, but a few specific signs call for a doctor.
Ibanez says slower speech, longer pauses, flatter emotional tone, word-finding difficulty, and vaguer language can all be subtle. They also vary widely from person to person and aren’t unique to aging or dementia.
Separately, Alrassi also told Everyday Health that hoarseness lasting more than three to four weeks, a lump in the neck, or trouble swallowing warrants medical evaluation. That is clinical advice, not something this study tested.
The researchers did not test whether changing the way someone speaks can slow aging. In other words, the interesting part is what speech may reveal, not whether people can game the clock.
Why Scientists Keep Chasing Speech
Collecting speech is cheap and non-invasive, which is much of the appeal.
Unlike brain scans or blood tests, researchers can record speech almost anywhere. Researchers can also repeat a short recording over time, giving them a relatively simple way to look for changes without relying on expensive equipment.
That could matter most in places where advanced diagnostic technology is harder to access. If the findings continue to hold across different populations and languages, speech could eventually complement other tools used to study healthy and accelerated aging.
Another question, though, may be even more interesting. Researchers now want to know whether the speech clock can move in the other direction. If someone’s brain health improves through effective interventions, would their speech begin to reflect that change too?
For now, nobody knows. But that possibility helps explain why something as ordinary as a conversation has become such an intriguing source of information about aging.
If a short voice recording could someday offer a clue about brain health, who should get to hear it: your doctor, a researcher, or only you?
More Articles You Might Be Interested in:
- 8 Creative Activities That Might Be Hurting Your Brain’s Aging Process
- Aerobic Exercise May Improve Brain Health in Postmenopausal Women, New Study Finds
- What Happens to Your Brain During Anesthesia? Scientists Are Still Searching for Answers
