How Accurate Are Biological Age Tests?
Reliability, age prediction and health-outcome prediction are three different questions.
AI-assisted editorial draft · Source check: 6 September 2026 · Independent reviewer: not yet assigned
Accuracy needs a defined target
A model trained to reproduce chronological age should be judged against that task. A model intended to estimate pace or associate with future outcomes needs different validation. One impressive accuracy statistic cannot answer all three questions.
Source 1: Genome Biology / HorvathSource 2: eLife / Belsky et al.Source 3: Aging / Lu et al.Ask about repeatability
Would two samples collected under comparable conditions give similar results? Ask the provider about technical variation, sample failures and the size of a change that exceeds normal noise. Results may also depend on the tissue and analysis pipeline.
Look beyond the headline
Check whether validation concerns the exact algorithm sold today, whether the people studied resemble the intended users, and whether an independent group replicated meaningful outcomes. A study linking a score to risk is not proof that lowering it changes that risk.
Sources & verification
- DNA methylation age of human tissues and cell types (2013)
Genome Biology / Horvath · Accessed 2026-09-06
- DunedinPACE, a DNA methylation biomarker of the pace of aging (2022)
eLife / Belsky et al. · Accessed 2026-09-06
- DNA methylation GrimAge strongly predicts lifespan and healthspan (2019)
Aging / Lu et al. · Accessed 2026-09-06