How accurate is Apple Watch VO2 max? About 6 points below a lab test

How accurate is Apple Watch VO2 max? In a 2025 validation study, 30 adults wore the watch for five to ten days to let it settle on an estimate, then ran a maximal treadmill test wearing a mask that measured the oxygen in every breath. The watch came in about 6 points below the laboratory figure on average [3].

The average hides how far a single reading can be off. For any one person, the watch's number sat somewhere between roughly 6 points above the lab result and 18 points below it [3]. The typical miss, in either direction, was close to 7 points - about 13% of the real value [3].

The laboratory number is not one fixed thing either, because the machine you are tested on changes it. A study of 30 athletes and 120 untrained people matched for age and sex, all between 17 and 25, tested both groups on a treadmill and on a stationary bike [1]:

• male athletes averaged 52.4 running and 45.2 pedalling
• female athletes averaged 41.0 running and 34.3 pedalling
• untrained men scored much the same either way, 33.4 and 34.2

Trained people give up about 7 points by being tested on a bike instead of a treadmill, while untrained people lose almost nothing [1]. Same body, different answer.

Home VO2 max tests that stop short of all-out effort carry their own bias. A 1985 study put 12 men aged 23 to 58 through four submaximal routines - easy pedalling and bench stepping, with the result extrapolated from how their heart rate climbed - and all four underestimated what the men actually measured [2]. The paper's own verdict was that the approach describes a group adequately, but that individual predictions are liable to considerable error [2].

Formulas that skip exercise altogether drift for a related reason: they only know what you tell them. Among 2,749 adults whose VO2 max was measured with breath analysis, a prediction built from age, body mass index - weight scaled to height - and an activity questionnaire had no slot for smoking, and heavy smokers, meaning more than 20 pack-years, or a pack a day for over 20 years, measured about 2.6 points below what those inputs implied [5]. Lighter smokers came in around 0.8 points below [5].

One method measures oxygen. The other guesses at it.

A laboratory test measures it directly. You exercise until you cannot continue, wearing a mask, and the equipment compares the oxygen in the air you breathe in with the oxygen in the air you breathe out. The difference is what your muscles took up. That technique is indirect calorimetry, which means working out energy use from the gases you exchange rather than from heat, and it is the reference the watch was judged against [3]. It is also how the treadmill and bike figures were produced [1].

A watch never sees your breath. The 2025 study let the device build its estimate from five to ten days of ordinary wear, then held that output up against the mask [3]. What gets tested is a prediction, and a prediction can only use what it already knows about you.

The smoking data shows what that costs. A formula fed age, body size and self-reported activity returns the same VO2 max for a heavy smoker and a non-smoker who answer identically, yet their measured capacity differs by about 2.6 points [5]. Whatever the model was never given ends up in the error rather than in the answer.

The wrist-estimate evidence here is one study, 30 adults, one brand, one protocol [3]. It compared a single watch reading against a single lab test, so it says nothing about whether the watch tracks your own change correctly over months.

Who this evidence does not cover:

• Garmin, Whoop, Fitbit, Polar and the rest - none of these papers tested them, and one manufacturer's algorithm tells you nothing about another's
• anyone outside 17 to 25, for the treadmill-versus-bike gap, which came from a single laboratory and one narrow age band [1]
• women, for the submaximal results, which came from 12 men in 1985 [2]

The smoking analysis is a group average across 2,749 people [5]. It cannot tell an individual smoker how far their own predicted figure is out.

None of this makes a wrist estimate worthless. It means the number arrives with a margin of error that the watch face does not display.

Suppose your watch has settled on 38. Applying the average gap from the validation study, a lab test would be expected to return something nearer 44 - though the spread in that study was wide enough that your real figure could also land below what the watch shows [3]. Comparing this month's watch reading with last month's is at least comparing one method against itself. Comparing it with a friend's treadmill result from a sports lab is not, and if that friend was tested on a bike instead, the two numbers drift further apart again [1]. The mask-and-treadmill test that clinics and university sports labs run is the thing your watch is estimating [3].