The Apple Watch VO2 max estimate got fitness percentiles right for 1 in 7 people

The VO2 max figure an Apple Watch shows is an estimate of maximal oxygen uptake inferred from heart rate and movement during ordinary walks and runs, not a measurement of it [s2]. A study putting the Apple Watch Series 10 against the laboratory criterion — analysing the gases a person actually breathes during a maximal exercise test — found it useful for sorting people roughly by fitness and unreliable as an exact number [s1].

The test

Forty healthy adults, half of them women, completed a maximal treadmill test with indirect calorimetry, the reference method, and then generated a VO2 max estimate from the watch [s1]. Data were collected between October 2024 and January 2025 [s1]. Agreement was assessed with Bland-Altman analysis alongside mean absolute error and mean absolute percentage error [s1].

Indirect calorimetry is the right comparison here. It captures the oxygen a person consumes and the carbon dioxide they produce at peak effort, which is what VO2 max is defined as. Anything a wrist device reports is a model trying to predict that value from signals correlated with it.

The gap is baked into how the estimate is made. Apple derives the figure from the relationship between heart rate and pace during ordinary outdoor walks and runs — submaximal efforts, mostly — and never observes the maximal exertion the laboratory test forces [s2]. It then extrapolates to a maximum the wearer may never reach. Layered onto that are the known error sources of the inputs: wrist optical heart rate drifts during exercise, GPS-derived pace is imperfect, and individual differences in running economy mean the same pace costs different people different amounts of oxygen. A model built on those signals can track the average population reasonably and still miss a specific person by a wide margin.

What it found

On average, the watch underestimated VO2 max by 6.25 mL/kg per minute (95% CI, −8.28 to −4.23), with limits of agreement running from −18.23 to 5.73 — a spread wide enough that an individual reading could be well off in either direction [s1]. The mean absolute error was 6.79 mL/kg per minute (95% CI, 4.91 to 8.67), and the mean absolute percentage error was 13.2% (95% CI, 9.6 to 16.8) [s1].

Translated into the fitness-percentile bands people actually see, the watch placed 5 of the 40 participants — 14% — in their correct band against a reference database, and a further 12 (34%) within one adjacent band [s1]. Roughly half, in other words, landed two or more bands from where a laboratory test put them. The authors' conclusion was that the estimates showed poor individual-level agreement [s1].

How to use the number

None of this makes the metric useless; it makes it a relative one. A systematic underestimate that is reasonably consistent within a person is still informative about change: if the figure climbs over months of training, that trend is probably real even when the absolute value is wrong. What it cannot support is a precise self-assessment — treating a single reading as one's true cardiorespiratory fitness, or comparing one's number against a friend's as if both were measured.

That distinction matters because VO2 max carries genuine weight. Across large cohorts, cardiorespiratory fitness is one of the stronger predictors of mortality — which is precisely why an inaccurate consumer estimate invites over-reading. The failure mode is the same one that afflicts wrist-based calorie estimates, which drift further from the truth as effort rises: a plausible number, delivered with false precision, that people act on as if it were measured. As a trend line, the watch's VO2 max earns its place. As a verdict on fitness, it does not.