What VO2max is, and what your watch estimates

VO2max is the most oxygen your body can use during hard exercise, in millilitres per kilogram of body weight per minute. Your watch doesn't measure that. It can't: it never sees your breath.

I don't test devices and I haven't worn any product named here. Each result names its authors, model and sample, and I checked all of this as of 3 November 2026.

A tracker isn't a medical device. Wearable readings are estimates for general fitness and wellness, and they don't diagnose or treat any condition. If you have symptoms or concerns, talk to a clinician.

What a lab does

In a typical lab test you exercise to exhaustion while a metabolic gas analyser reads your breathing. Caserman and colleagues used that method in their 2024 study, on a cycle ergometer, as the reference for a watch.

A watch takes a shortcut. It works from data it does have, chiefly heart rate and pace on your runs, and puts them through a formula you can't inspect. Some devices lean on resting data instead, and that split matters, as the numbers show.

How close the estimates get

The best overview I found is from INTERLIVE, a European research network that includes one technology company, Huawei. Molina-Garcia and colleagues (2022, peer-reviewed, Sports Medicine) pooled 14 validation studies. It's also manufacturer-funded, in part, by Huawei, and the literature is pre-2022.

The average looks wonderful. The limits of agreement are the part that applies to you. Exercise-based estimates were good for a group, but a single person could still land about 10 ml/kg/min high or low. On a true figure of 45, that's anywhere from about 35 to 55. In fitness terms, those are very different runners.

What brand studies found

These are small. They also disagree, and none transfers to another model or to newer firmware.

Apple Watch. Caserman 2024 (peer-reviewed, JMIR Biomed Eng) tested a Series 7 on 19 adults, 7 of them female, aged 18 to 63. The lab read 45.88 ml/kg/min and the watch 41.37, with a mean absolute percentage error of 15.79%. The lab test was on a bike and the watch's estimate came from an outdoor run, which may account for part of that gap. Lambe's 2026 review (peer-reviewed, authors declare no competing interests) cites one Apple Watch study of 30 people with a mean difference of -6.07 ml/kg/min and wide limits of agreement, and calls it clinically significant. That's one study, so don't read it as a verdict on the brand.

Garmin. Three studies, three different answers:

• Carrier 2025 (peer-reviewed, no external funding, Sensors), Fenix 6, 19 people: 7.05% error, concordance 0.73. Inside the 10% threshold those researchers used.
• Howard 2026 (peer-reviewed, abstract read, funding unchecked), Forerunner 265, 20 people: the watch read significantly higher than the lab, 43.95 against 41.10 ml/kg/min at the start. Error stayed under 10%, and the abstract gives no exact figure.
• Passler 2019 (peer-reviewed, no external funding, Int J Environ Res Public Health), Forerunner 920XT, 24 people: 7.3% error, reading significantly low. Other devices in the study passed 10%, and those authors concluded no device gave valid results.

So "Garmin is accurate" isn't a sentence these papers support. Nor is "Garmin is inaccurate". One read high, one read low, and each covers a single model.

I couldn't verify any figure for this year's watches or firmware. I found no independent test of them. Most papers don't report a firmware version anyway.

Trend, not a lab result

Here's how I'd read it. The absolute number is a rough guess about you. Its direction over months, on the same watch and similar routes, is more useful, because whatever error the formula makes in your case may stay roughly stable.

"May" is doing real work in that sentence. I didn't find a study that tests how well a watch follows one person's change over a long period. The nearest is Howard 2026: across 12 weeks of training, the Forerunner 265 went from 43.95 to 45.00 ml/kg/min and the lab from 41.10 to 42.43. Both moved up and the watch stayed high. That's 20 people, two time points and one model. It's encouraging and nothing more.

What this means for a marathon plan

Three habits, if you're following a plan:

1. Don't compare your number with a friend's on another brand. The formulas differ.
2. Don't panic over a drop of 1 or 2 points. That's far inside the error band above.
3. Don't set a goal pace from it.

The race predictor deserves the same caution. In Howard's study the Forerunner 265 predicted a half marathon of 21.1 km (13.1 mi) significantly too fast: 129.97 minutes against 137.55 minutes actually run. That's over 7 minutes, on one model, in 20 people, though the authors still put the prediction error under 10%. I'd take your pace from recent runs and your plan, and treat the predictor as a talking point.

Since the estimate leans on heart rate, a poor reading will feed through it, so getting an accurate heart rate from your watch is worth doing first. If you're choosing a watch for a race, try best watches for your first marathon or the first marathon picks.