In short
Our estimator starts at your chronological age and applies eight literature-anchored adjustments. Every weight is listed below. It is a lifestyle-proxy educational model, not an epigenetic clock, and it deliberately contains no product term.
Most online biological age calculators do not show their working. That is usually because the working is either arbitrary or designed to produce a result that sells something. Ours is published in full so you can disagree with it specifically.
How the model works
The estimator begins at your chronological age and adds or subtracts years across eight domains. The adjustments are additive and independent, which is a simplification: in reality these factors interact, and someone with poor sleep and poor fitness likely carries more combined risk than adding the two terms suggests. We accept that simplification for legibility.
The output is bounded. It will not return a figure more than twelve years below your chronological age, more than twenty above, or below eighteen. Real clocks produce wider spreads. The bound is there to stop a stacked set of answers from returning a number that would mislead more than it informs.
The complete weighting table
| Domain | Response | Years applied | Anchor |
|---|---|---|---|
| Tobacco | Never smoked | 0.0 | Large pooled cohort analyses put long-term smoking at roughly a decade of life expectancy. We apply approximately half that as a biological age proxy. |
| Quit 10+ years ago | +0.4 | ||
| Quit within 10 years | +2.0 | ||
| Current smoker | +5.5 | ||
| Physical activity | 300+ min a week | −2.5 | Graded dose-response between cardiorespiratory fitness and all-cause mortality observed in large treadmill-tested cohorts, with no upper benefit threshold found. |
| 150 to 300 min | −1.4 | ||
| Around 150 min | 0.0 | ||
| Under 150 min | +2.6 | ||
| Central adiposity | Waist-to-height < 0.45 | −1.0 | Waist-to-height ratio outperforms BMI as a cardiometabolic risk screen; 0.5 is the conventional boundary. |
| 0.45 to 0.50 | 0.0 | ||
| 0.50 to 0.55 | +1.8 | ||
| Above 0.55 | +3.4 | ||
| Sleep | 7 to 8.5 h, restorative | −1.0 | Meta-analysis across more than a million participants shows a U-shaped duration-mortality relationship, elevated at both tails. |
| 6 to 7 h or inconsistent | +1.0 | ||
| Under 6 h or insomnia | +2.8 | ||
| Habitually over 9 h | +1.2 | ||
| Resting heart rate | Under 60 bpm | −1.2 | Cohort meta-analysis associates each 10 bpm increment with a meaningful rise in all-cause mortality. |
| 60 to 70 bpm | 0.0 | ||
| 71 to 80 bpm | +1.0 | ||
| Above 80 bpm | +2.2 | ||
| Alcohol | Little or none | 0.0 | Weighted toward more recent analyses that control for former-drinker bias, which removed much of the apparent protective effect at low intake. |
| Up to 7 a week | +0.3 | ||
| 8 to 14 a week | +1.2 | ||
| Over 14 a week | +2.8 | ||
| Diet quality | Whole-food dominant | −1.5 | Dietary pattern research is consistent at pattern level even where single-nutrient findings conflict. |
| Mixed | 0.0 | ||
| Mostly ultra-processed | +2.0 | ||
| Psychosocial stress | Low, well managed | −0.8 | Chronic unrecovered stress is associated with inflammatory and metabolic marker changes. Weighted conservatively because self-report is unreliable here. |
| Moderate | 0.0 | ||
| Chronically high | +1.8 |
What this model deliberately does not include
Any product term. There is no input for supplement use and no path by which taking anything changes your result. That is a design decision. Producing a projected age reduction from a product purchase would require effect-size evidence that does not exist, and a calculator that invents one is an advertisement wearing a lab coat.
Genetics. Heritable factors contribute meaningfully and we cannot capture them from eight questions.
Clinical biomarkers. The best-validated composite ageing measures use blood panels. Levine's PhenoAge, for instance, uses nine clinical markers including albumin, creatinine, glucose, C-reactive protein, lymphocyte percentage, mean cell volume, red cell distribution width, alkaline phosphatase and white cell count. We use lifestyle proxies because most people do not have those numbers to hand. Proxies are weaker.
Disease history and medication. Both matter and both are outside the scope of a public educational tool.
Known limitations
- All inputs are self-reported, and self-report is systematically optimistic on diet, activity and alcohol.
- Additive independence is a simplification; the real relationships interact.
- The weights are drawn from population averages and may not apply to you individually.
- Most of the underlying cohorts skew toward European ancestry populations, which limits generalisation.
- It is not validated against methylation clocks, because it is not attempting to replicate one.
What to use instead if you want precision
A standard blood panel from your doctor gives you real biomarkers and costs little. If you want an actual epigenetic measure, DNA methylation testing is commercially available, though the clocks disagree with each other and test-retest variability on a single sample is larger than most providers advertise. Treat one methylation result as a data point, not a verdict.
Once you have your estimate, the useful next read is which levers actually move it.
Common questions
Is this the same as an epigenetic clock?
No. Epigenetic clocks measure DNA methylation at specific genomic sites from a biological sample. This is a lifestyle-proxy model built from questionnaire inputs and published cohort associations. It is educational and directional, not a laboratory measure.
Why does the calculator not include supplements?
Because including one would require a defensible effect size for a supplement's impact on biological age, and that number does not exist in the literature. A calculator that invents it is marketing material, not a tool. We would rather ship a smaller honest model.
Can I use this to track progress over time?
Yes, with a caveat. Re-run it no more often than every ninety days and answer consistently. Because the model is banded, a small real change may not move your result at all, and crossing a band boundary produces a step change larger than the underlying improvement. Treat direction as the signal.
Why is the result bounded?
Stacking every worst-case answer would otherwise return a figure that is not informative. Bounding the output keeps extreme combinations inside a range where the estimate still means something.