Wearable VO₂ Max Estimates

The "VO₂ max" or "Cardio Fitness" number on your watch is a poor measurement that may still work as a trend tracker. It is probably least accurate for the people most likely to watch it — the very fit, and on thinner evidence the very unfit — but the direction it drifts over months may be the more honest part, though that is barely tested, and it is the part worth using.

Read the watch's figure as a trend over months, not a measurement. Give it one steady, flat outdoor run or walk a week, never use the absolute number for medical decisions or training zones, and if the real figure matters, anchor it with one lab test.

Your watch is guessing, not measuring

A watch does not measure oxygen uptake. Measuring VO₂ max — the most oxygen your body can use — takes a mask, an all-out effort and analysis of your breath in a lab.[1] Instead, it fits a line of heart rate against pace during an ordinary, below-maximum effort, using its optical sensor and GPS, then extends that line to your age-predicted maximum heart rate.[2]

Apple updates from a flat outdoor walk, run or hike that lifts your heart rate to at least 30% of your heart-rate reserve (the gap between resting and maximum), using your age, sex and weight; you can flag heart-rate-lowering drugs such as beta-blockers. If you take one and don't flag it, the watch will overestimate your fitness.[3] Garmin needs an outdoor run of at least ten minutes at moderate intensity.[4]

Fine on average, unreliable for you

Pooled across eight studies and 244 people, the average error was close to zero, but an individual reading could land about 10 mL/kg/min either side of the truth.[5] The one study to run a strict equivalence test found the watch failed it.[6] These are small studies, mostly 12 to 44 people, so the pattern is reliable but the precise figures are rough.

DeviceWhat validation found
Apple Watch Series 9 / Ultra 2Read 6.1 mL/kg/min low on average (a 12% shortfall); individual readings from about 6 too high to 18 too low[7]
Apple Watch Series 10Error unchanged a generation later; only 5 of 35 people put in the right fitness percentile[8]
Garmin Vivoactive 4s (unfit group)Read about 14 mL/kg/min high, closer to a formula from age, height and weight than to the lab[9]
Polar V800Accurate in women, not in men; large random error, up to about 16 mL/kg/min, across the two watches tested[10]

Brand matters too. When the same 24 people wore two watches, one read about 3 mL/kg/min high and the other about 2 low, so your trend line doesn't survive switching watches.[11]

Apart from one conference study in unfit adults with a median age in the mid-50s, the watch studies used young, mostly fit volunteers (no mean age above 32). The 2025 Apple Watch study excluded anyone on heart-rate-affecting medication, exactly who the beta-blocker setting is for. And several checked running-based estimates against a bike test, which an expert consortium advises against.[12]

Probably most wrong for the very fit, and maybe the very unfit

Across the small studies so far, watches probably pull everyone toward the average, overestimating unfit people and underestimating fit ones. Tested against a bike test, an Apple Watch Series 7 read the least-fit group about 4 mL/kg/min high and the fittest group about 12 low — an error near 20%. Those groups held only three and five people, so trust the direction more than the size.[13] The newest Apple Watch, tested mostly in young, fit people, read low on average and lower in the fitter group; it had too few unfit people to test the other half.[14] Garmin's engine shows the same squeeze: small errors in moderately trained users, wider ones in the highly trained.[15]

The likely reason is that the algorithms are anchored to population averages, above all the age-predicted maximum heart rate. The vendors document the anchoring; that it causes this pattern is plausible, not measured.

Midlife adults of middling fitness may fare best. In 74 adults aged 35–64 with high blood pressure, prediabetes or metabolic syndrome, the estimate was essentially unbiased with a typical error around 10%, worse only in type 2 diabetes. That used a chest-worn sensor, not a wrist reading, and no wrist validation exists in people over 65.[16]

What makes the number jump around

Anything that raises heart rate without raising oxygen demand reads as a fake fitness change. Each of these has a well-established effect on heart rate, but few have been measured against watch VO₂ directly:

  • Heat. Blood goes to the skin to cool you, heart rate climbs, and the watch reads lost fitness.
  • Illness, poor sleep, stress, caffeine. All raise exercise heart rate and push the estimate down on the day; Apple names heat and caffeine as causes of underestimation.[17]
  • Cold. Narrowed blood vessels weaken the pulse signal at the wrist that the whole model depends on.[18]
  • Skin tone. Darker skin absorbs more of the sensor's green light, one of the better-evidenced sources of error.[19] Average wrist pulse error doesn't differ by skin tone, but the spread of errors is about twice as wide in dark skin,[20] and it shows up during exercise — around 12 beats per minute of error above 60% of heart-rate reserve, the very range the estimate is computed in.[21]
  • A loose strap or hard arm-swing can make the sensor lock onto your stride instead of your pulse.[22]
  • Stale body weight. The figure is per kilogram, so a wrong weight gives a wrong number.
  • The software. A firmware update can freeze or change the metric with nothing changing in your body. Unlike the ECG features on the same watches, this feature has no FDA clearance from either Apple or Garmin.

The trend may be the more honest part — with a blind spot

Worn the same way on similar routes, the watch's bias should stay roughly constant, so the change over months probably says more than the number — but the evidence for that is thin. In a large population study that re-tested about 2,000 adults some seven years apart, the change predicted by a research chest sensor, not a watch, tracked the change in treadmill-estimated fitness only moderately — a correlation of about 0.57, where 1.0 would be perfect.[23]

Apple has measured repeatability once: a typical user's readings scattered by about 1.2 mL/kg/min, but more than twice that for one user in ten, a sizeable fraction of the bias itself.[24] Nobody has repeated it independently, and Garmin has no equivalent.

The models were better at detecting improvement than decline, an observation rather than a tested comparison. Trust a rising trend, look into a sustained fall, and don't read a flat line as proof that nothing is slipping.

Don't let it set your training zones

Set intensity by how you feel and breathe. Apple and Garmin build heart-rate zones from maximum heart rate rather than the VO₂ estimate, but the watch estimates that too and can get it wrong. The VO₂ figure does feed Garmin's race predictor and suggested workouts: an underestimate sets your suggested Zone 2 paces too slow; an overestimate in a beginner can push "easy" runs into anaerobic territory.

Getting a cleaner reading

You can't make it lab-accurate, but you can cut the avoidable error:

  • Use a chest strap (such as a Polar H10) on runs for cleaner heart rate, as for heart-rate variability. It won't fix the anchoring, the bigger problem.
  • Wear it snug, just above the wrist bone.
  • Give it one steady, flat outdoor run a week; the systematic review suggests 10–15 minutes is enough.
  • Compare like with like across similar seasons, and keep your weight current in the app.
  • If you're highly trained and thinking in treadmill-test terms, mentally add 10–20%. The right correction depends on test and watch: nearer 28% against a bike test, while one Garmin group was already within 7%, so adding anything would overshoot.
  • Anchor it once with a lab test if you train enough to care, though lab numbers also vary with the test used. See VO₂ max for what that involves.

Further reading

  • Lambe R et al. Investigating the accuracy of Apple Watch VO₂ max measurements: A validation study. PLoS One 2025.[25]
  • Lambe R et al. Accuracy of VO₂ max Estimates From Apple Watch Series 10. Mayo Clin Proc Digit Health 2026.[26]
  • Caserman P et al. Assessing the Accuracy of Smartwatch-Based Estimation of Maximum Oxygen Uptake Using the Apple Watch Series 7: Validation Study. JMIR Biomed Eng 2024.[27]
  • Molina-Garcia P et al. Validity of Estimating the Maximal Oxygen Consumption by Consumer Wearables: A Systematic Review with Meta-analysis and Expert Statement of the INTERLIVE Network. Sports Med 2022.[28]
  • Železnik Mežan I. Accuracy of wearables for determining the maximal oxygen uptake and lactate threshold: a qualitative systematic review. Front Sports Act Living 2025.[29]
  • Carrier B et al. Validation of Aerobic Capacity (VO₂max) and Pulse Oximetry in Wearable Technology (Garmin Fenix 6). Sensors (Basel) 2025.[30]
  • Passler S et al. Validity of Wrist-Worn Activity Trackers for Estimating VO₂max and Energy Expenditure (Garmin Forerunner 920XT, Polar V800). Int J Environ Res Public Health 2019.[31]
  • Spathis D et al. Longitudinal cardio-respiratory fitness prediction through wearables in free-living environments — the Fenland Study. NPJ Digit Med 2022.[32]
  • Rissanen AE et al. Cardiorespiratory Fitness Estimation Based on Heart Rate and Body Acceleration in Adults With Cardiovascular Risk Factors: Validation Study. JMIR Cardio 2022.[33]
  • Singh B et al. Impact of Skin Pigmentation on Pulse Oximetry Blood Oxygenation and Wearable Pulse Rate Accuracy: Systematic Review and Meta-Analysis. J Med Internet Res 2024.[34]
  • Hung SH et al. Validity of heart rate measurements in wrist-based monitors across skin tones during exercise. PLoS One 2025.[35]

— § —