How Ageon works
Every number in Ageon, how it's worked out and the study behind it. Most apps keep their models to themselves. We'd rather you could check ours.
The Ageon Score
Each measurement becomes a percentile against people of your age and sex, from the published table named next to it. A domain is the weighted average of its measurements, and the score is the weighted average of the six domains. A domain you haven't measured drops out, and the others scale up to fill it.
Tables marked verified copy their paper's numbers, and tests pin a few of them so a typo fails the build. Tables marked draft approximate their source, for example from a mean and spread, and get checked as real results come in.
Cardio
30%- VO2 maxFRIEND registry treadmill tests, Kaminsky et al., Mayo Clinic Proceedings, 2022verified
- Blood pressureWhelton et al., ACC/AHA guideline, Hypertension, 2018; Prospective Studies Collaboration, Lancet, 2002draft
- Resting heart rateNHANES 1999 to 2008, Ostchega et al., National Health Statistics Reports no. 41, 2011verified
- Heart rate variabilityVoss et al., Short-term heart rate variability: influence of gender and age in healthy subjects, PLoS One, 2015draft
- Cardio recoveryCole et al., New England Journal of Medicine, 1999; Qiu et al., Journal of the American Heart Association, 2017draft
Strength
20%- Push-upsCSEP Physical Activity, Fitness and Lifestyle Approach rating bands, 2003 (female table adapted to full push-ups)draft
- Sit-to-riseAraújo et al., Sitting-rising test: sex- and age-reference scores derived from 6141 adults, European Journal of Preventive Cardiology, 2020verified
- Chair standsMcKay et al., Neurology, 2017 (ages 20 to 59); Rikli and Jones, Journal of Aging and Physical Activity, 1999 (60 and over)draft
- Grip strengthDodds et al., Grip strength across the life course: normative data from twelve British studies, PLoS One, 2014verified
Sleep
15%- Sleep durationNational Sleep Foundation recommendations, Hirshkowitz et al., Sleep Health, 2015draft
- Sleep consistencyHuang and Redline, Diabetes Care, 2019; Windred et al., Sleep, 2024draft
- Sleep midpointSaint-Maurice et al., Sleep, 2024draft
- Deep sleepOhayon et al., Sleep, 2004; Robbins et al., Sensors, 2024draft
- REM sleepOhayon et al., Sleep, 2004; Leary et al., JAMA Neurology, 2020; Robbins et al., Sensors, 2024draft
- Time awake at nightLichstein et al., Behaviour Research and Therapy, 2003; Saint-Maurice et al., Sleep, 2024; Robbins et al., Sensors, 2024draft
Activity
15%- Daily stepsNHANES 2005 to 2006 accelerometer data, Tudor-Locke et al., Medicine & Science in Sports & Exercise, 2009draft
- Weekly exerciseWHO guidelines on physical activity and sedentary behaviour, Bull et al., British Journal of Sports Medicine, 2020draft
Body composition
10%- Waist-to-height ratioAshwell et al., Obesity Reviews, 2012draft
- Body fatGallagher et al., American Journal of Clinical Nutrition, 2000draft
- Lean mass indexKyle et al., Nutrition, 2003; Cederholm et al., ESPEN consensus, Clinical Nutrition, 2015draft
Mobility and balance
10%- One-leg balanceAraújo et al., Successful 10-second one-legged stance performance predicts survival, British Journal of Sports Medicine, 2022draft
- Walking speedObuchi et al., Reference value on daily living walking parameters among Japanese adults, Geriatrics & Gerontology International, 2020draft
Ageon Age
Your Ageon Age is the age at which your numbers would score exactly average. Ageon tries your results against the norms for other ages until it finds the one where you'd sit at the 50th percentile. It stays within 15 years of your real age, because beyond that the tables get too thin to say more.
Only measurements whose norms change with age can date you: VO2 max, heart rate variability, push-ups, sit-to-rise, grip strength, one-leg balance and daily steps. Sleep, waist, blood pressure and the rest are scored against fixed health ranges, so they move your score but not your age. The age you see updates once a week, so a bad night doesn't move it.
Estimated or Verified
A score is Verified when cardio comes from a watch or a lab, and sleep and activity from a wearable. Anything resting on a questionnaire or a self-test is Estimated, and each domain says which it is. You don't need a watch: the tests, a timed 5K and a few questions give a full score, marked as estimated.
Apple Watch reads VO2 max about 6 mL/kg/min low against a lab treadmill test (Lambe et al., PLoS One, 2025), so its reading is scored at 1.1 times. A run or 5K estimate is never adjusted.
Pace of Aging
Pace is how many years your Ageon Age moves per calendar year, from a line fitted through the last 26 weeks. 1× means aging like the calendar, and under 1× is slower. A noisy few weeks are pulled toward 1× in proportion to how uncertain the line is, so the pace only claims what the data shows.
Blood age
Blood age is PhenoAge (Levine et al., Aging, 2018), from nine markers on a routine blood test: albumin, creatinine, glucose, c-reactive protein, lymphocytes, mean cell volume, red cell distribution width, alkaline phosphatase and white blood cells. It needs all nine from the same draw. Nothing is estimated or carried over from another test. It sits next to Ageon Age and never changes it, because most people test once or twice a year.
Each marker is shown against a cited adult range in ink, never red. Reports you upload are read by Anthropic's Claude to fill in the values, and you check every one before anything is saved.
Experiments
An experiment compares your numbers during a change with the 4 weeks before it. Ageon only calls a change clear when it's at least 3 standard errors from your normal ups and downs, and possible at 2. It also has to be big enough to matter:
- HRV: at least 3 ms
- Resting heart rate: at least 1.5 bpm
- Sleep: at least 15 minutes
- Steps: at least 750 steps
The bar sits higher than the usual 2 because daily numbers are correlated and each experiment watches several at once. A false "it worked" costs you more than a missed one, so Ageon says "no clear change" often. That's an honest answer, and it tells you what isn't your lever.
What we don't know yet
- Draft tables approximate their papers. They'll be checked against real results and marked verified.
- Some tables were measured differently from how a phone or watch measures: seated pulse against a watch's resting heart rate, 5 minutes of HRV lying down against readings through the day, steps from a hip counter against a wrist.
- The age-group comparison assumes domains correlate at 0.3 until there are enough people per age band to use the real spread.
Scoring version 2026.10.10. Ageon is a wellness estimate, not medical advice. See also the Learn articles, each with its studies.