Your watch's VO2max is a decent estimate for ordinary fitness, shakier at the elite end

The VO2max figure a running watch shows — an estimate of maximal oxygen uptake, the standard measure of cardiorespiratory fitness — is more trustworthy than most of the numbers a wearable produces, at least for ordinary fitness levels. Two systematic reviews in early 2026 found consumer devices give valid or acceptable VO2max estimates in most studies of untrained and recreational exercisers, while cautioning that accuracy frays at the elite end and that the estimates carry wide, study-dependent variability [s1][s2].

Cardiorespiratory fitness is worth estimating well: it is one of the stronger predictors of long-term health. The appeal of a wearable estimate is that it substitutes a submaximal run in the real world for a laboratory test to exhaustion — a genuinely useful trade if the number holds up.

What the reviews found

A qualitative systematic review examined 13 studies, most of them testing Garmin smartwatches paired with a chest-belt heart-rate sensor and the Firstbeat Technologies algorithm [s1]. In every case, the reference VO2max was measured in the laboratory with a graded treadmill test, while the device generated its estimate from submaximal outdoor runs [s1]. Seven of the studies found the wearable estimate valid or acceptable against the gold standard, and three demonstrated valid estimation of the lactate threshold, a related endurance marker [s1].

The review's nuance is about who it works for. Valid estimates were found across healthy untrained adults, recreational athletes and team-sport professionals — but usefulness in elite endurance sport was questionable and may depend on more advanced algorithms [s1]. A practical finding was that using two or more submaximal runs as the input, rather than one, could improve validity [s1]. The reassuring headline is that a good estimate did not require a maximal effort — the submaximal run was enough [s1].

The variability underneath the average

A separate systematic review and meta-analysis looked specifically at estimating cardiorespiratory fitness from free-living, unsupervised wearable data — the more ambitious version, in which the device infers fitness from everyday movement rather than a structured test [s2]. Pooling 18 studies covering 31,072 participants, it found a promising overall correlation of 0.83 (95% CI 0.77–0.88) between predicted and measured fitness [s2].

But the heterogeneity was very high — an I² of 97% — meaning the studies disagreed enough that a single pooled number understates how much individual results scatter [s2]. The authors flagged most concerns in the data-analysis domain and a general lack of external validation, and concluded that no firm verdict on clinical implementation can be drawn yet [s2]. This is the same reproducibility gap that runs through wearables in clinical research more broadly: a strong average correlation that has not been shown to transfer between devices and populations.

How the estimate is actually made

It helps to know what the watch is doing. The Firstbeat approach that dominated the reviewed studies does not measure oxygen at all; it infers maximal capacity from the relationship between heart rate and pace or power during submaximal effort, then extrapolates to what your uptake would be at maximum [s1]. That is why an accurate heart-rate input is essential, and why the reviewers singled out feeding the algorithm two or more submaximal runs as a way to sharpen the estimate — more data points anchor the extrapolation [s1]. It also explains the elite-athlete caveat: in highly trained people the heart-rate-to-workload relationship is compressed and idiosyncratic, so a model tuned on the general population extrapolates less well [s1].

The free-living meta-analysis was more ambitious still, drawing on four databases and pooling results with a Fisher Z transformation across 18 studies, eight of which used more advanced machine-learning models [s2]. Its participants had a weighted mean age of 46.9 years, and its central caution was methodological: the risk-of-bias concerns clustered in how studies handled and reported their data, and external validation was largely absent [s2]. A pooled correlation of 0.83 built on that base is a reason for optimism about the concept, not yet a licence to treat a passively estimated fitness number as clinically reliable [s2].

What it means for a user

For tracking your own fitness over time, a watch's VO2max estimate is one of the more defensible numbers it offers — especially if you are an untrained or recreational exerciser, feed it consistent submaximal runs, and use a device-and-sensor combination that has been validated [s1]. Read it as a personal trend line: a rising estimate over months is more meaningful than the exact figure, and a comparison against another person on another device is not reliable given the scatter the meta-analysis exposed [s2]. At the sharp end of endurance performance, where small differences matter most, the laboratory test still earns its place [s1]. The estimate depends heavily on an accurate heart-rate input, which itself varies by device during exercise.