Machine learning tool predicts biological aging from voice patterns

Researchers developed an AI system that analyzes hundreds of speech characteristics to estimate a person's biological age, potentially identifying dementia risk. In a study of nearly 3,000 Spanish-speaking adults, individuals whose estimated vocal age exceeded their chronological age showed higher rates of cognitive decline, dementia, and biological aging acceleration. This non-invasive approach offers advantages over existing aging clocks that require expensive brain scans or blood tests.

The research team analyzed speech samples from participants across five Latin American countries, examining over 700 distinct vocal characteristics ranging from speaking pace and pitch to vocabulary complexity and emotional tone. Participants engaged in standardized speaking tasks designed to capture natural language patterns, from describing visual content to rapid word generation and delayed story recall. The machine-learning model learned to associate these vocal features with chronological age, then compared its predictions against actual ages to identify individuals whose speech patterns suggested accelerated aging.

The study revealed important socioeconomic dimensions alongside health findings. Those with older-sounding voices relative to their actual age not only demonstrated higher dementia and cognitive decline rates but also tended to have less favorable social and economic circumstances, suggesting potential links between life conditions, aging acceleration, and cognitive health outcomes.