Biological aging clocks don't measure what we think they measure

Biological aging clocks—tools built with machine learning on molecular data—aspire to measure biological age, that nebulous concept of rising mortality risk independent of chronological years. A recent narrative review classifies these instruments into three generations: the first optimized to predict chronological age, the second designed to estimate morbidity and mortality risk, and the third quantifying the pace of aging from longitudinal biomarker changes. The mechanistic framework rests on methylome remodeling, immunosenescence, and chronic low-grade inflammation, integrating multi-omics biomarkers and artificial intelligence approaches to capture aging's multidimensional nature. Yet critical limitations persist: questionable biological specificity, reverse causality, statistical confounding, lack of generalization across populations, and unproven clinical applicability. The sobering takeaway is this: each clock represents a distinct operationalization of biological aging, not interchangeable measures of a single construct, so readers genuinely interested in longevity must demand mechanistic longitudinal studies in diverse populations before trusting any individual prediction.