How Accurate Are Smartwatch Calories? Typical Errors of 15 to 25%, and Worse for People With More Body Fat

Smartwatch calories are an estimate, and usually an overestimate. In a study published in July 2026, four popular watches were compared with a lab metabolic system. Typical errors were about 15% to 25%, three of the four watches significantly overestimated calories burned, and the error grew with body fat for every brand [1]. The authors concluded that current devices “do not yet provide reliable caloric monitoring for individuals or for research” [1].

That matches a decade of research. Watches have become good at heart rate and steps. Calories remain the least accurate number they report [2][3]. This post covers what the new study found, why calories are so hard, and what that means for anyone who eats, coaches or builds products around the number.

What did the 2026 study find?

Jason Kostrna and colleagues at Florida International University tested 58 adults, all Hispanic, with Fitzpatrick skin types III to V and a mean BMI of 29.7 [1]. Each did a single session on a recumbent bike: five minutes of rest, ten minutes alternating moderate and vigorous intervals, then five minutes of recovery. They wore four watches at once, and a COSMED K5 metabolic system measured the energy they actually used:

• Apple Watch Series 8
• Garmin Forerunner 955
• Samsung Galaxy Watch 5
• Fitbit Sense 2

Four findings stand out:

1. Three of four overestimated. Average bias was +21.6 kcal for Apple, +56.8 for Samsung and +68.6 for Garmin over a session of about 20 minutes, all statistically significant [1].
2. Fitbit’s clean result needs an asterisk. Its average bias of +3.1 kcal was not significant, but only after seven readings more than four and a half times the true value were removed as implausible. With them, Fitbit’s average bias was +128.6 kcal [1].
3. Error rose with body fat, for every brand. The study found a significant effect of body fat percentage and a significant interaction between device and body fat [1]. The people most likely to be counting calories to lose weight got the least accurate counts.
4. No skin tone effect, within the range tested. Skin type made no measurable difference to calorie error across Fitzpatrick III to V [1]. The study had no participants with the lightest or darkest skin, so it cannot rule out an effect there. Skin tone matters more for optical heart rate during exercise, which we covered in wearable accuracy on darker skin.

The lead author’s advice to users was to round the number down and pay more attention to heart rate, heart rate zones and their own trends over time [4].

Is this new?

No. Every major evaluation has found the same pattern: good heart rate, poor calories.

• Stanford, 2017. Sixty volunteers wore seven wrist devices while sitting, walking, running and cycling. Six of seven measured heart rate within 5%. For energy expenditure, the most accurate device was off by about 27% and the least accurate by 93%, and no device was below 20%. The Apple Watch had the lowest error on both [2].
• Apple Watch meta-analysis, 2025. A University of Mississippi team pooled 56 studies of the Apple Watch and found mean absolute percentage errors of 4.43% for heart rate, 8.17% for steps and 27.96% for energy expenditure, with the calorie error consistent across activities and populations [3].

The 2026 study adds the body fat finding and a current set of devices, but the headline has not moved in nearly a decade.

Why are calories so hard to get right?

Because a watch does not measure them. A lab measures energy from the oxygen you consume and the carbon dioxide you breathe out. A watch has no way to do that, so it estimates from things it can measure or is told [5][6]:

• Heart rate, from the optical sensor;
• Motion, from the accelerometer, and pace and distance from GPS;
• Your profile: Apple lists height, weight, sex and age as inputs [6].

Each brand turns those into calories with its own proprietary equations, which is why devices disagree with each other as well as with the lab [2]. Heart rate is a good proxy for effort but an imperfect one for energy: it also rises with heat, stress, caffeine and dehydration, and the relationship between heart rate and energy differs between people.

There is a second, quieter estimate inside the daily total. Resting energy, the calories your body uses just to keep running, is not measured at all. It is calculated from profile data [6]. An app’s “total calories” is active plus resting, so an error in either carries straight into the total.

Where does the error do damage?

Food budgets. Many nutrition apps add exercise calories back into the day’s allowance. MyFitnessPal, for example, adds calories from manually logged or synced workouts to the daily burn and raises the food goal straight away, and can add a separate adjustment when a tracker’s full-day estimate beats its own, according to a MyFitnessPal staff explanation [9]. An overestimated workout becomes permission to eat food that was never burned.

The arithmetic is unforgiving. In the illustration above, a watch that reads 25% high on five 500 kcal workouts a week hands back 500 kcal a week that was never spent. Over a year that is about 26,000 kcal of food the plan assumed had been burned.

Weight loss and GLP-1 programs. Programs built around energy balance inherit the same problem, and the study suggests it is worst for users with more body fat [1]. Calorie burn is a weak basis for coaching decisions in these programs; we covered what they measure instead in GLP-1 digital weight management.

Cross-device comparisons. A Garmin user and an Apple Watch user doing the same workout can see calorie numbers 40 or 50 kcal apart on the evidence above [1]. Leaderboards, challenges and rewards built on calories reward the device, not the effort.

How should the number be used?

For people wearing the watch:

• Round down, and treat the figure as an upper estimate [4].
• Compare yourself with yourself, on the same device, over weeks. Trends are more trustworthy than single sessions.
• Lean on heart rate and time in zone for training decisions; both are measured far more accurately [2][3]. Our guide to heart rate zones explains why zones still differ by vendor.
• Keep your profile current and calibrate. Apple recommends updating height, weight, sex and age and calibrating with about 20 minutes of outdoor walking or running with GPS [6][10]. It improves the inputs, not the underlying estimate.

For teams building on calorie data:

• Do not add exercise calories back at face value. Offer a partial credit, or none, and say why.
• Store the source with every value, and never compare calories across brands as if they were the same measurement. See deduplicating health data across wearables.
• Keep active and resting energy separate in storage, and only add them when both come from the same source [7][8].
• Prefer measured signals for rewards. Steps, active minutes and heart rate based intensity are harder to inflate than calories.
• Be careful with users with higher body fat, who are both the most likely to rely on the number and the most likely to get a wrong one [1].

The short version

A July 2026 study found typical smartwatch calorie errors of about 15% to 25%, significant overestimates from Apple, Samsung and Garmin, and larger errors for people with more body fat [1]. That fits Stanford’s 2017 finding that no device got below 20% error and a 2025 meta-analysis putting Apple Watch calorie error at 28% against 4% for heart rate [2][3]. Calories are estimated from heart rate, motion and profile data, and resting calories are not measured at all [5][6]. Use the number as an upper estimate and a trend, keep active and resting energy separate, and be wary of any product that hands it back as food.