How accurate are wearable calorie counts?

Calories are the weakest number on your wrist. A device can give you 2,437 kcal to the single calorie, and the studies below say that precision is decoration.

I don't test devices and I haven't worn any product named here. Each result names its authors, model and year, and I checked 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're changing what you eat for a health reason, or you have symptoms or concerns, talk to a clinician.

The short version

Don't rely on a wearable's calorie figure to plan what you eat. Every large review I read ends in about the same place.

• Fuller 2020 (peer-reviewed, JMIR mHealth): a review of 158 publications across 9 brands. Its wording is flat: "For energy expenditure, no brand was accurate." The literature ends in May 2019, and one author joined Garmin after the paper was finished.
• Shcherbina 2017 (peer-reviewed, J Pers Med): 60 volunteers, age 38 ± 11, on Apple Watch, Basis Peak, Fitbit Surge, Microsoft Band, Mio Alpha 2, PulseOn and Samsung Gear S2. No device got calorie error below 20%. The same paper's heart-rate results were far better, so a good pulse reading doesn't buy you a good calorie figure. Most of those devices no longer exist.
• Lambe 2026 (peer-reviewed, authors declare no competing interests): 8 Apple Watch studies, 270 participants. Error ran from 9.71% in running to 151.66% in walking, and all six studies reporting it hit 20% or more in at least one condition. Five of the 8 used Series 2 or older, and the authors couldn't pool the results.

Look at that Apple Watch range again. Same brand, 9.71% to 151.66%, depending on what you were doing and which study you read.

Why it varies so much

A watch doesn't measure calories. It estimates them from motion and heart rate, using a formula you can't inspect.

Adding heart rate does help. O'Driscoll and colleagues (2020, peer-reviewed, Br J Sports Med; abstract read, funding unchecked) pooled 60 studies and 104 effect sizes, and adding heart rate to motion data cut the error in most activity types. Accuracy still depended heavily on the activity, and results varied so widely that heterogeneity passed 75% for many devices.

So the same watch can do reasonably on a run and badly on a walk or in the gym.

Weight training is the worst case

Lee and colleagues (2026, peer-reviewed, Sensors; funded by SOLUM Health Care, which employs one author) tested Apple Watch Series 7, Galaxy Watch 4, Fitbit Charge 5 and Garmin Vivosmart 5 on 62 healthy men with a mean age of 26.55. In resistance exercise, the correlation with the lab measure was r = 0.10 to 0.34, with reliability (ICC) under 0.45. Endurance sessions were underestimated, with wide limits of agreement. It's the newest study I found, and it's men only.

Two more, both measuring how far readings scatter (coefficient of variation):

Both are peer-reviewed, and both used young, healthy samples. One Düking author had Polar funding on an unrelated earlier project. I haven't checked funding for Hajj-Boutros.

Fitbit tends to read low

This is the one brand-wide direction I can give you, and it's dated. Chevance and colleagues (2022, peer-reviewed, JMIR mHealth) reviewed 52 studies and meta-analysed 41. With high-risk-of-bias studies removed, Fitbit under-read by 2.77 kcal per minute on average, with limits of agreement of -12.75 to 7.41. "Recent" there means Fitbit models as of 2022, so it doesn't tell you about the Fitbit Edge or Air.

That spread matters more than the average. One person in one session can land a long way from -2.77.

What about current watches?

There's a gap here, and I'll say so plainly. Lee 2026 is as recent as I can offer, and its watches are a generation or more old. I found no independent calorie test of this year's models or firmware. Makers can change the formula through an update, so an old result doesn't follow a watch forever, in either direction.

Be wary of anyone ranking the "most accurate calorie tracker". INTERLIVE, a European research network with Huawei among its members, published a best-practice protocol for this kind of testing (Argent 2022, part-funded by Huawei), so there's a yardstick to ask about.

How I'd use the number

My view, and it's only that: use it to compare you against you. If this week's figure is higher than last month's and you've been walking more, the direction may be useful. I wouldn't eat back an exercise figure on the strength of it. A 20% miss on a large day is a meal.

Things that count well are better habit anchors. Fuller's review found Fitbit, Apple Watch and Samsung appeared to measure steps accurately in the lab, for instance, so a step target is a firmer thing to chase than a calorie target.

That's the reasoning behind my list of trackers for weight loss, which goes by habit. The weight loss picks start from the same place. Calorie estimates lean on your heart-rate reading too, so how optical heart rate works is the next useful read.