Biological Age Tests: Do Epigenetic Clocks Actually Work?

The one-sentence version

Epigenetic clocks are a genuine scientific achievement — they predict mortality and disease risk across large populations better than chance — and they are also too noisy, too unstandardized, and too unvalidated to tell you as an individual whether a supplement, a diet, or a clinic visit actually changed how fast you are aging. Both halves are true at once. This page holds them together, because the longevity industry sells you the first half and quietly omits the second. For where biological-age testing sits in the wider toolkit, see our pillar on longevity medicine: what's proven vs hyped.

What a "biological age test" actually measures

Your chronological age is how many birthdays you've had. Your biological age is meant to capture how worn your body is — the idea being that two 50-year-olds can be aging at very different rates. Most consumer "biological age" tests estimate this from DNA methylation: chemical tags (methyl groups) that sit on your DNA at specific sites called CpGs and change in patterned ways as you age. An algorithm reads the methylation pattern at hundreds of these sites and outputs a single number — your "epigenetic age" — usually from a blood or saliva sample. (A handful of tests skip methylation entirely and read a different aging signal — for example the inflammation-tracking antibody-sugar test we grade in our GlycanAge review.)

The algorithms that do this are called epigenetic clocks. They are not all the same, and the differences matter enormously when you're deciding whether to trust one.

The clocks, briefly — and why generation matters

The first clocks were trained to predict chronological age. Horvath's 2013 multi-tissue clock used 353 CpG sites and estimated age across many tissue types with striking accuracy¹. The Hannum 2013 clock did something similar in blood². These "first-generation" clocks answer the question "how old does this methylation pattern look?" — which, paradoxically, makes them less useful for health, because a clock that perfectly predicts your birthday tells you nothing your driver's license doesn't.

The breakthrough was realizing that the errors were the interesting part. A clock that estimates you a few years "older" than your birthday — a positive age acceleration — turns out to flag people at higher risk. DNAm age predicts all-cause mortality in later life: in a landmark analysis pooling multiple cohorts, people whose methylation age ran ahead of their chronological age died sooner, even after adjusting for known risk factors³.

That insight produced the second-generation clocks, trained directly on health and mortality rather than on age:

• •PhenoAge (Levine 2018) was built to predict a composite of clinical aging measures and outperformed first-generation clocks at predicting lifespan and healthspan⁴.
• •GrimAge (Lu 2019) was trained on mortality and surrogate biomarkers of smoking and inflammation; it is among the strongest single predictors of time-to-death and time-to-disease in the literature⁵.

Then came a different idea entirely. DunedinPACE (Belsky 2022) doesn't estimate an age at all — it estimates your pace of aging, calibrated against decades of longitudinal organ-system decline in a single birth cohort followed since 1972. It reports a rate: 1.0 means you're aging one biological year per chronological year; 1.2 means 20% faster⁶. Conceptually it's the most appealing for tracking change over time.

Here's the catch the marketing skips: these clocks disagree with each other. They were trained on different data toward different targets, so the same blood sample can return a "younger" GrimAge and an "older" PhenoAge. There is no single canonical "your biological age." (We break down which clock is built for which job — mortality, disease, or pace of aging — in our head-to-head GrimAge vs PhenoAge vs DunedinPACE comparison.) When a DTC test hands you one headline number, it has made an invisible choice about which clock to trust — and you usually aren't told which, or why.

Where epigenetic clocks genuinely work

Give the science its due. In population research, the best second-generation clocks are real, reproducible tools:

• •They predict mortality and morbidity. GrimAge and PhenoAge consistently associate with time-to-death, cardiovascular events, and age-related disease across independent cohorts⁴⁵.
• •They track known biology. Clock acceleration moves in the expected direction with smoking, obesity, and chronic disease — the patterns aren't random.
• •They have become a standard endpoint in aging research. A 2023 framework on biomarkers of aging positions methylation clocks among the most developed candidate measures for evaluating longevity interventions in trials⁷, and a 2025 expert consensus statement treats them as legitimate — but explicitly research-stage — tools for intervention studies⁸.

If you are running a randomized trial in thousands of people, an epigenetic clock is a defensible way to ask whether a treatment nudged the average rate of aging. That is the use case where these tools shine.

Where they break down — for *you*

The problem is that "works as a population research endpoint" and "works as a personal health readout" are completely different bars. Three failures separate them.

1. They're noisy at the individual level

The single biggest issue is measurement reliability. The same DNA sample, run twice, can yield epigenetic-age estimates that differ by several years — not because you aged between draws, but because of technical variation in the methylation assay. A detailed reliability study found that many widely used clocks have poor test–retest reproducibility, with first-generation clocks especially unstable⁹. When the noise is measured in years and the change you're hoping to detect from a supplement is a fraction of a year, the signal drowns. A single reading can swing you "two years younger" with zero change in your actual biology.

This isn't a fringe critique — it's well enough established that researchers built a fix. Principal-component (PC) clocks (Higgins-Chen 2022) were engineered specifically to bolster reliability for clinical trials and longitudinal tracking, dramatically reducing the technical noise of the original clocks¹⁰. The fact that the field had to re-engineer the clocks to make them trustworthy enough for repeated measurement tells you how unreliable the consumer versions — which generally don't use PC methods — can be.

2. They're not standardized

There is no FDA-cleared "biological age" diagnostic. Different companies use different clocks, different assays, different normalization, and different reference populations. Two reputable tests can hand the same person meaningfully different ages, and there's no regulatory floor guaranteeing either is right. The expert consensus is explicit that aging biomarkers, including methylation clocks, still need standardization and validation before clinical deployment⁸.

3. The thing you actually care about is unproven

Here is the question that matters: if I lower my epigenetic age, do I live longer or healthier? That has not been shown. Clocks are validated as predictors (acceleration associates with worse outcomes), not as modifiable targets (proven that pushing the number down extends your life). It is entirely possible for an intervention to lower a clock reading without lowering your actual risk — the clock would be a marker the treatment moved without moving the underlying disease. Demonstrating that a clock is a valid surrogate requires showing that changing it changes hard outcomes, and that work largely hasn't been done. This is the same "raises a lab value vs. improves your health" gap we flag for the whole field in longevity medicine: what's proven vs hyped.

"But this study reversed my epigenetic age!"

You'll see this claim a lot. The most-cited example is a 2021 pilot randomized trial of an 8-week diet-and-lifestyle program that reported a roughly 3-year reduction in Horvath clock age versus controls¹¹. It's a real, peer-reviewed result — and it's also exactly the kind of study to read carefully:

• •It was small and short (a few dozen men, 8 weeks) — a pilot, by the authors' own framing.
• •It used a first-generation clock (Horvath), which the reliability literature flags as among the noisiest⁹.
• •It measured a change in a biomarker, not a change in any health outcome. No one lived longer in 8 weeks; the clock simply read lower.

A reduction in a noisy first-generation clock over 8 weeks in a small pilot is hypothesis-generating, not proof that the participants aged backward. Treat every "we reversed your epigenetic age" claim — especially from a company selling the supplement and the test together — as marketing until a large, well-controlled trial using reliable clocks and hard endpoints replicates it. The viral "8 years younger" supplement claim is a textbook case of exactly this trap — we dissect how a placebo-free, manufacturer-linked study used an unvalidated clock to sell it in alpha-ketoglutarate (Rejuvant): does the "8 years younger" claim hold up?. The same caution applies to the "about 2.5 years younger" biological-age headline behind the fasting-mimicking diet (ProLon), which rests on aging-clock surrogates and company-affiliated research.

So should you buy one?

A measured take:

• •As a one-time curiosity or a rough population-style risk signal: fine, if you can afford it and won't over-read the number. A markedly accelerated GrimAge in a heavy smoker is telling you something you probably already knew.
• •As a tool to track whether your supplement stack or clinic protocol is "working": no. The test-retest noise is large enough that you'll see "improvements" that are pure measurement variation, and the clocks aren't validated as modifiable targets anyway. You'll spend money confirming nothing.
• •As a substitute for treating actual risk factors: never. The boring, proven levers — blood pressure, LDL, glucose, smoking, fitness, sleep — have hard-outcome evidence that no epigenetic clock can match.

If you do want a biological-age readout, the best-built consumer option is worth picking carefully — we grade the most prominent DTC methylation test, which reports the DunedinPACE clock, in our TruDiagnostic TruAge review — and the founder-branded membership version of TruAge in our Tally Health review. And understand that the providers selling it mostly test but don't treat — the defining catch of the DTC lab band. We map that trade-off in longevity clinics vs lab memberships vs Rx telehealth, and we sort which markers on a longevity panel are actually useful versus vanity in what do longevity biomarker panels actually test?. And before paying for an epigenetic clock at all, see whether a free biological-age test or calculator gets you a more actionable number — the open, outcome-validated PhenoAge formula runs off a basic blood panel for nothing. For a free, lab-free measure that predicts mortality through strength and balance, see the sitting-rising test and longevity.

Bottom line

Epigenetic clocks are a legitimate scientific tool that predicts aging-related risk across populations — and a poor personal dashboard. They disagree with each other, they're noisy enough that a single sample can swing by years, they aren't standardized or FDA-cleared, and — most important — lowering your clock score has never been shown to make you live longer. Use them, if at all, as a one-time curiosity, not as a scoreboard for your supplement budget. If you've decided to buy one anyway, we rank the at-home options by defensibility in our best at-home biological-age test guide. For an independently graded look at the providers selling biological-age testing and the longevity therapies around it, see our longevity clinic rankings.