Apple Watch VO2 Max Test: What Apple Changed
Last Updated: September 21, 2026
Apple announced a guided at-home VO2 max test on September 9, 2026. It uses the iPhone camera alongside a heart rate device, gives real-time audio and visual feedback, and returns a result on the spot. It has not shipped. It sits in the iOS 27.2 beta, and Apple says it arrives later this year with the redesigned Health app. Until then, Apple Watch still estimates VO2 max the way it always has, passively, from outdoor walks, runs and hikes. Three independent studies agree that estimate runs low, by roughly 13 to 16 percent. Here is what changes, what does not, and what the peer-reviewed data actually shows.
Want to compare your own device against the lab data first? Try the VO2 max accuracy tool, which ranks nine wearables against lab CPET results, then come back for the story behind the numbers.
I have been tracking wearable VO2 max validation for about a year, and this is the first time a major brand has changed the measurement itself rather than tuning the algorithm behind it. That matters more than it sounds, because every accuracy figure in this post was measured against the old method.
What Apple actually announced
Two things landed on September 9, and it is easy to run them together.
The hardware. Apple Watch Series 12 and Ultra 4 shipped September 18 with what Apple calls a new Health Sensing System: larger green LEDs, a larger ECG electrode surface, heart rate sampled every five seconds all day, and heart rate variability captured up to 24 times more often than before. Apple says this is the most accurate heart rate sensing in a wearable, based on its own study of more than 1,000 participants against ECG chest straps. None of that changes how VO2 max is calculated.
The software. Separately, Apple previewed a redesigned Health app built around Apple Intelligence. Inside it sits the new Apple Watch VO2 max test: a guided physical assessment run at home using the iPhone camera, with your heart rate supplied by an Apple Watch, AirPods Pro 3, or a third-party heart rate sensor. Apple says the camera work runs entirely on device and no video is ever recorded, stored, or shared.
The same release brings Readiness, a 0 to 10 morning score built from recent activity, sleep and vitals, and Health Age, which blends VO2 max, resting heart rate, HRV and sleep with blood biomarkers to compare your metrics against your chronological age.
The timing is the part most coverage gets wrong. The redesigned Health app is being tested in the iOS 27.2 beta, but reporting from MacRumors says it will not be in the 27.2 launch build. Apple's own wording is "later this year." So the test exists, and nobody outside Apple has used it in a finished form.
How Apple Watch measures VO2 max today
Everything below describes the method that is live on your wrist right now.
Apple Watch Series 3 and later records a VO2 max estimate during an Outdoor Walk, Outdoor Run, or Hiking workout. Indoor workouts do not count. The estimate comes from heart rate against GPS pace, adjusted by your profile, with an assumed maximum heart rate filling the gap that no wrist sensor can measure. Apple's support documentation puts the supported range at 14 to 65 mL/kg/min.
That range is worth a second look. Apple's own 2021 healthcare whitepaper described the supported range as 14 to 60. The current support page says 14 to 65. The ceiling moved, and Apple never published a note explaining when or why.
Apple does publish an accuracy claim, which is easy to miss because it lives in a healthcare PDF rather than a support page. In that whitepaper, Apple reports a mean error of 1.2 mL/kg/min with a standard deviation of 4.4, and an intraclass correlation of 0.89, across a design cohort of 534 people and a validation cohort of 221.
Read that carefully. Apple is claiming the average person's estimate lands within about one unit of a lab test. Three independent studies disagree.
What the independent research says
Device | Sample | Finding | Study |
Series 10 | 35 adults | Underestimates by 6.25 mL/kg/min, MAPE 13.2% | Lambe et al. 2026, Mayo Clinic Proceedings: Digital Health |
Series 9 or Ultra 2 | 28 adults | Underestimates by 6.07 mL/kg/min, MAPE 13.3% | Lambe et al. 2025, PLOS ONE |
Series 7 | 19 adults | MAPE 15.8%, predicted about 41 vs measured about 46 | Caserman et al. 2024, JMIR Biomedical Engineering |
Three studies, three research groups, one direction. Apple Watch reads low, and the size of the gap has barely moved across three generations of hardware.
The 2026 Series 10 study is the newest and the most detailed. Participants ran a modified Astrand treadmill protocol against a COSMED metabolic cart. Apple Watch underestimated by 6.25 mL/kg/min, with a 95 percent confidence interval of negative 8.28 to negative 4.23. Mean absolute error was 6.79. Mean absolute percentage error was 13.2 percent.
Set those numbers against Apple's claimed mean error of 1.2, and the gap is roughly five units of VO2 max. For context, five units is about the difference between the 40th and 60th percentile for cardiorespiratory fitness in a middle-aged adult.
The number that matters more than MAPE
Buried in the Series 10 paper is a statistic that reframes the whole topic. Only 5 of 35 participants, 14 percent, were placed in the correct cardiorespiratory fitness percentile band. Another 12, or 34 percent, landed one band away. That means more than half of the group was misclassified by at least one fitness category.
Percentage error is easy to dismiss. Thirteen percent sounds tolerable. Being told you are average when you are above average, or fit when you are not, is a different kind of wrong, and it is the kind that matters if you are using the number to make decisions about training or health.
The limits of agreement tell the same story: negative 18.23 to positive 5.73 mL/kg/min. The group average is off by six. Any individual person can be off by eighteen.
Where Apple sits against the other brands
The independent evidence base is thinner than most buyers assume. Only four brands have non-vendor peer-reviewed VO2 max validation at all.
Brand | Method | Independent validation | Direction |
Garmin (Firstbeat) | Outdoor run, HR to pace | Yes, strongest | MAPE around 7%, underestimates highly trained runners |
Apple | Outdoor walk, run or hike | Yes, three studies | Underestimates by about 6 mL/kg/min |
Polar | Resting Fitness Test or run test | Yes | Slight underestimate vs CPET, MAPE 13.7% |
Fitbit / Google | 10 minute outdoor GPS run | Charge 2 era only | Overestimated, and the tested algorithm has been retired |
Samsung | Outdoor run | No, heart rate only | Unverified for VO2 max |
WHOOP | GPS run, proprietary model | No | Company data only |
Oura | In-app guided walk test | No | Vendor states it is less accurate than a lab test |
Coros | Outdoor run, HR to pace | No | No published figures |
Suunto | Shared Firstbeat engine | No brand specific study | Inferred from Garmin |
The pattern across all of them is set by method rather than brand. The INTERLIVE meta-analysis found exercise-based estimates carry a group-level bias of negative 0.09 mL/kg/min, essentially nil, while resting-based estimates read high by 2.17. Individual error stays wide either way, with limits of agreement around plus or minus 13 to 17 for resting methods.
If you want the full ranking with every study and its funding source, the most accurate VO2 max wearable breakdown covers all nine devices in detail.
Your VO2 max number is only as good as what you eat and sleep
VO2 max is not a fixed trait. It moves with training load, sleep, iron status, hydration and recovery, and the VO2 max factor explorer ranks more than 30 of those inputs by evidence strength. The catch is that a watch shows you the output without showing you which input moved it.
That is the gap Kygo closes. Log your food in 20 to 30 seconds by voice, photo, text or barcode, and Kygo runs correlation analysis across 12 to 36 hour windows against the biometric data already flowing in from Apple Health, Garmin, Oura, Fitbit and WHOOP. Instead of watching a number drift, you see which of your own patterns moved with it.
What the new test could fix, and what it will not
A guided test solves a real problem. The current estimate needs an outdoor GPS workout at a reasonable intensity, and it silently skips workouts that do not qualify. Apple's own whitepaper reported that estimates were produced for 79 percent of outdoor pedestrian workouts longer than 5.75 minutes. If you train indoors, lift, cycle, or walk gently, your number can sit stale for weeks. An on-demand test removes that dependency entirely.
It also removes the pace assumption. Estimating from HR against GPS pace means terrain, wind, urban GPS drift and running economy all leak into the result. A structured protocol with real-time feedback controls the effort instead of inferring it.
Three things it almost certainly does not fix.
Assumed maximum heart rate. Nothing in a camera or an optical sensor measures your true max HR. If the model still leans on an age-based estimate, the largest single error source survives the redesign. This is the input flagged in both the INTERLIVE analysis and Firstbeat's own methodology as the main driver of error in highly trained people.
The extremes. Every device in this category underestimates in highly trained people and overestimates in sedentary ones. That is a property of population-fitted models, not of the sensor, and a new protocol does not automatically change the fit.
It is still not a CPET. A lab test analyzes the oxygen and carbon dioxide in your breath. No watch, ring, phone camera or earbud does gas analysis. Every consumer number on this page is an inference.
The precedent nobody is talking about
Fitbit already ran this experiment, and the result should temper expectations.
Fitbit's cardio fitness score was validated twice, both times on the Charge 2, both times in 2019. Google then retired the resting heart rate path in May 2026, leaving a method that requires a 10 minute outdoor GPS run. The two studies that validated Fitbit validated an algorithm that no longer exists, and no validation of the current method has been published. We covered what that broke for owners in why Fitbit Air says VO2 max is not tracked.
The same reset is about to hit Apple. The moment the guided test ships, the three studies in this post describe a method that is no longer the primary one. Independent validation of a consumer wearable typically lands two to four years after the hardware, which is why most validation studies in this field test devices that are already a generation or two old. Apple Watch VO2 max is about to become, on paper, the best-studied estimate in the category and simultaneously an unvalidated one.
That is not a knock on Apple. It is the cost of changing a method, and it is why the WHOOP situation is worth reading as a contrast: WHOOP VO2 max accuracy rests entirely on the company's own data, and the company's own data says the error is worst at the extremes.
What to do with your number right now
Four things, in order of how much they change the result.
Track the trend, not the value. Reliability is much better than validity here. Apple's whitepaper reports an intraclass correlation of 0.86 to 0.89, meaning the number is consistent with itself even when it is several units away from a lab result. A four-week direction of travel is informative. The absolute figure is not.
Pair a chest strap for hard efforts. Wrist optical heart rate tracks well at steady effort and degrades during hard or variable work, which is exactly when VO2 max is being estimated. This single change removes the largest correctable error.
Give it the workout it wants. Outdoor, GPS on, a sustained effort above an easy stroll, on open ground rather than under buildings or tree cover. Indoor sessions produce nothing.
Assume you are reading low. Three studies agree on the direction. If your Apple Watch says 42 and a Garmin says 47, the research says the Garmin is probably closer, and the same person's lab value is likely higher than both.
Common questions
Is the Apple Watch VO2 max test available now?
No. Apple announced it on September 9, 2026 and it has not shipped. It is being tested in the iOS 27.2 beta, but reporting indicates it will not be in the 27.2 launch build, and Apple says the redesigned Health app arrives later this year.
What do you need to take the Apple Watch VO2 max test?
An iPhone with a compatible camera, plus a heart rate source. Apple lists Apple Watch, AirPods Pro 3, or a third-party heart rate sensing device. The camera portion runs on device and no video is recorded or stored.
Does the Apple Watch underestimate VO2 max?
Yes. Three independent studies all found underestimation: 6.07 mL/kg/min with a MAPE of 13.3 percent (Lambe 2025), a MAPE of 15.8 percent on the Series 7 (Caserman 2024), and 6.25 mL/kg/min with a MAPE of 13.2 percent on the Series 10 (Lambe 2026).
How accurate is Apple Watch VO2 max compared to Garmin?
Garmin has the strongest independent validation, at roughly 7 percent error in general populations, against Apple's 13 to 16 percent across three studies. Both underestimate in highly trained people. Garmin's advantage narrows when the Apple Watch user wears a chest strap.
Why did my VO2 max change after a software update?
Apple has adjusted the estimate without documenting the change. The supported range moved from 14 to 60 mL/kg/min in Apple's 2021 whitepaper to 14 to 65 on the current support page, and users reported score shifts after the iOS 26 update in September 2025. Apple published no explanation.
Will the new test be more accurate?
There is no independent data yet, and there will not be until researchers can run the shipped version against a metabolic cart. A structured protocol should reduce the noise from GPS pace and terrain. It does not obviously fix the assumed maximum heart rate, which is the largest known error source.
Can Apple Watch VO2 max replace a lab test?
No. A lab CPET analyzes the oxygen and carbon dioxide in your breath. The Series 10 study placed only 14 percent of participants in the correct fitness percentile, so the estimate is not a substitute for a clinical measurement.
What workouts give Apple Watch a VO2 max reading?
Outdoor Walk, Outdoor Run and Hiking, on Apple Watch Series 3 or later. Indoor workouts do not produce an estimate, which is why many users see the number go stale.
The bottom line
Apple is replacing a passive estimate with a real test, which is the most interesting thing any brand has done in this category in years. It is also not here yet. The number on your wrist today comes from the same outdoor-workout method three independent studies have measured, and all three say it reads about six units low, with more than half of people landing in the wrong fitness percentile.
Treat the figure as a trend line rather than a score. Wear a chest strap when you want the estimate to be good. And hold judgment on the new test until somebody outside Apple puts it against a metabolic cart, which on the historical pattern will take a couple of years.
Kygo connects the food side to the biometric side so you can see which of your own habits move the metrics your watch reports. Start on iOS or Android.
Key sources: Lambe et al. 2025 (PLOS ONE, Apple Watch validation); Lambe et al. 2026 (Mayo Clinic Proceedings: Digital Health, Series 10); Caserman et al. 2024 (JMIR Biomedical Engineering, Series 7); Apple's cardio fitness whitepaper (May 2021, vendor claim); Apple support documentation (current supported range); Apple Newsroom (September 2026 announcement); Molina-Garcia et al. 2022 (Sports Medicine, INTERLIVE meta-analysis); Carrier et al. 2025 (Sensors, Garmin); Neudorfer et al. 2025 (Sensors, Polar); Freeberg et al. 2019 and Klepin et al. 2019 (Fitbit Charge 2).
A note on sources. Ship dates and beta contents for the redesigned Health app come from press reporting, not from an Apple release note, and Apple has published no protocol or accuracy figure for the new test. The frequently repeated "Galaxy Watch plus or minus 4.7 mL/kg/min" figure appears only on marketing and AI-generated content sites and could not be traced to an indexed study, so it is excluded here. Apple's own accuracy figures are vendor-reported and are labelled as such throughout.
Disclaimer: Kygo is a personal data aggregation and insights platform designed for informational purposes only. The information provided by Kygo, including correlations, patterns, and trends identified in your data, does not constitute medical advice, diagnosis, or treatment. Always consult a licensed healthcare provider with any questions regarding medical conditions.
If your Apple Watch VO2 max ever disagreed sharply with a lab test or another device, I would like to hear which way it went and by how much.