Epigenetic clocks estimate biological age by measuring DNA methylation, chemical tags attached to DNA that regulate gene expression, at specific sites across the genome. As people age, methylation patterns at these sites shift in ways that are statistically predictable enough to build an age-estimating algorithm from, one that in many studies tracks health outcomes and mortality risk better than chronological age alone.
The field has moved through several generations of clocks. The original Horvath clock estimated chronological age directly. Later clocks like GrimAge and DunedinPACE were trained instead to predict mortality risk and pace of aging, which turned out to correlate more usefully with actual health outcomes than simply predicting someone's birth year. This is the most extensively validated biomarker category in longevity science, published across hundreds of peer-reviewed studies, which is why it carries a strong evidence rating relative to newer or more speculative measures.
Epigenetic age remains a population-level statistical estimate, not a diagnostic test for an individual, and results can vary meaningfully between different clock algorithms run on the same blood sample. It's best understood as a research and trend-tracking tool rather than a single definitive number.