How accurate are smartwatch calories? What a 2026 lab study found across four brands

Short answer

Smartwatch calorie counts are rough estimates, not measurements. In a 2026 lab study of 58 adults, Apple Watch, Garmin, Samsung and Fitbit mostly overestimated calories burned during a short cycling test, with typical errors of about 15% to 25%. The higher a person's body fat, the larger the error for every brand.

The study, published in PLOS ONE by researchers at Florida International University, compared each watch with indirect calorimetry, the lab method that calculates energy use from the oxygen you breathe in and the carbon dioxide you breathe out (1). Garmin and Samsung overestimated the most on average, Apple less, and Fitbit's average was close to zero only after a batch of implausible readings was removed. Skin tone did not change accuracy in this group.

This does not mean your watch is useless or that its heart rate is wrong. Calories are the one number the watch has to calculate from other signals, while heart rate is read more directly from the sensor. Earlier reviews of more than 150 studies reached the same conclusion: heart rate and steps hold up reasonably well in the lab, calories do not (4).

If you have been eating "back" the calories your watch says you burned and the scale has not moved, the watch's estimate may be part of the reason, and it is not a failure of willpower. Before you blame yourself, check what your body already recorded: your heart rate, sleep and activity trends are often more useful than the calorie total.

Note: this article is for general education and is not medical advice. No consumer wearable measures calories directly; every device estimates them with its own algorithm. Do not use a watch's calorie number to set a strict eating target, and talk to a doctor about unexplained weight change. Welltory is a general wellness product and does not diagnose any condition.

What did the 2026 smartwatch calorie study actually test?

The study is titled "Body fat, skin tone, and the accuracy of smartwatch caloric expenditure estimates." It was published on July 29, 2026, by Jason Kostrna, Ekaterina Oparina, Cristina Palacios, Andres J. Rodriguez, JunZhu Pei, Ajmal Ajmal and Jessica C. Ramella-Roman (1). The university press release followed on September 1, 2026, and the story was picked up widely in September (2).

Here is what the researchers did, in plain terms:

• Who took part: 58 Hispanic adults aged 18 to 50, 31 women and 27 men. The average age was 23. Average BMI was 29.7; 25 participants had a BMI of 30 or above.

• Which watches: Apple Watch Series 8, Garmin Forerunner 955, Samsung Galaxy Watch5 and Fitbit Sense 2. Each person wore the watches at the same time, with wrist positions shuffled between participants.

• What they did: five minutes of rest, then about 10 minutes of cycling on a recumbent bike in two-minute blocks that alternated between moderate and vigorous effort, then five minutes of recovery. The watches ran their indoor cycling mode for the whole session.

• What the watches were compared with: a portable metabolic system (COSMED K5) that analyzes every breath, plus a chest-strap heart rate monitor.

• What else was measured: body fat percentage (by bioimpedance and skinfold calipers) and skin tone on the Fitzpatrick scale, types III to V.

That design makes the study useful, and it also sets clear limits on what it can tell you. It looked at one type of exercise, done once, for a short time, in a lab, in one fairly young and fairly homogeneous group. It did not test running, walking, weight training, team sports or a full day of normal life. It also tested 2022-generation watches, and companies update their algorithms over time.

So the honest reading is narrow: during a short, controlled cycling session, these four popular watches got calories meaningfully wrong, and they got them more wrong for people with more body fat.

How far off were Apple Watch, Garmin, Samsung and Fitbit?

The paper reports error in two ways, and it helps to keep them apart.

1. Average bias in kilocalories. This is how much the watch over- or underestimated on average, for the whole session, compared with the lab measurement. Positive numbers mean the watch said you burned more than you did (1):

Three of the four watches showed a statistically significant average overestimate. Fitbit's average was not significantly different from zero, but that result comes with a big asterisk: seven Fitbit sessions, about 13% of its trials, produced estimates more than four and a half times the lab value, and the researchers removed them as implausible. With those sessions included, Fitbit's average overestimate rose to 128.6 kcal, the largest of any device. The press release also notes that one watch reported just one calorie for an entire workout (2).

2. Absolute percentage error. This ignores the direction of the miss and asks how big it was relative to the real value. The paper reports that median absolute percentage error was "consistently high," ranging from about 15% to 25% across devices, and that the average absolute miss was roughly 80 to 110 kcal depending on the watch (1). The FIU press release uses the same 15% to 25% range for "typical errors" and adds that averages for the worst-performing watches "climbed much higher" (2).

Some coverage of the study has quoted errors of 50% to 100% for people with high body fat. We could not find those figures in the paper or in the university's press release, so we do not repeat them here. What the paper does show is a clear trend: the more body fat, the bigger the percentage error.

Two more details matter for reading the table:

• The numbers are for one short session. The watches ran for about 20 minutes, including rest and recovery. A 60 kcal overestimate on a session like that is not small: lead author Jason Kostrna told FIU News that "you could be off hundreds of calories each week" (2). But you cannot multiply the table by your weekly minutes and get a precise answer, because error differs by activity and by person.

• The brand order is not a ranking. After the researchers adjusted for body fat in a statistical model, the estimated errors shifted, especially for Fitbit, and the authors attribute part of that gap to the removed outlier sessions. Earlier studies have also put different brands at the top and bottom. These are 2022 models in one type of exercise. The safest takeaway is that all four watches had substantial calorie error, not that one brand is good and another is bad.

Why does body fat make calorie estimates worse?

This is the most important new finding. In the researchers' statistical models, body fat percentage had a significant effect on error, and the size of that effect differed by brand. Error went up with body fat for all four watches, with steeper increases for Fitbit and Garmin than for Apple in percentage terms, and for Garmin and Samsung in absolute kilocalories (1).

The authors are careful not to claim they know why. Watch makers do not publish their calorie formulas, so the researchers write that the proprietary nature of the algorithms "prevents mechanistic attribution of error patterns." What they do suggest is that algorithms need to be trained on more diverse data, grouped by body fat level, to work equally well across body types. Kostrna made the same point to FIU News: "When the companies are developing those algorithms, we don't know where the test data is coming from" (2).

It helps to know what a watch knows about your body. Usually it is height, weight, age and sex from your profile. It does not know how much of your weight is fat and how much is muscle, and muscle uses more energy than fat. Two people with the same weight and the same heart rate during a ride can burn different amounts of energy. That is a plausible reason a weight-based model might miss more for some bodies than others, but it is an explanation, not something this study proved.

The practical point is uncomfortable but useful. The people most likely to be tracking calories for weight loss may be the people for whom the calorie estimate is least reliable. If that is you, the number on your wrist is not a judgment of your effort. It is a model that was not built around your body.

Why didn't skin tone affect accuracy?

Optical heart rate sensors shine light into the skin and read how much comes back, and earlier work raised concern that skin tone could make wearables less accurate for some people. The 2017 Stanford study of seven wrist devices found device error was higher with darker skin tone, among other factors (3).

In the 2026 study, Fitzpatrick skin type III to V had no meaningful effect on calorie error (1). Two caveats keep this from being the final word. Only four participants had type V skin, so the study had little power to detect a difference there, and no one had type VI. The researchers also tried two instrument-based skin tone measurements and found them unreliable in their setup, so they relied on self-reported skin type.

So the fair summary is: in this group, body fat mattered and skin tone did not. It does not settle the question for all skin tones or for heart rate.

Why can't a smartwatch measure calories?

A watch never sees calories. The lab method in this study captures the gases you breathe, because burning fuel uses oxygen and produces carbon dioxide in predictable amounts. A watch on your wrist has no access to your breath.

What a typical smartwatch has is:

• An optical heart rate sensor that reads your pulse from blood flow under the skin.

• An accelerometer (and often a gyroscope) that reads how your wrist moves.

• Your profile: height, weight, age and sex.

The paper describes it the same way: most wrist wearables "combine photoplethysmography, accelerometry, and proprietary algorithms to estimate caloric expenditure" (1). The algorithm turns those inputs into a calorie estimate. Your total calories add an estimate of what your body burns at rest, usually calculated from your profile with a standard equation.

That creates a ladder of how close each number sits to the sensor:

1. Heart rate is closest. The sensor reads your pulse and reports it. It can still be wrong, especially with a loose strap or fast wrist movement, but it is a direct reading.

2. Heart rate variability (HRV) is one step further. It is calculated from the timing between beats, so it depends on clean data, but it comes from your own heart signal.

3. Steps are a pattern-recognition problem: the watch decides which wrist movements are steps. Our article on how fitness trackers count steps explains where that goes wrong.

4. Calories are furthest away. They are a prediction built on top of heart rate, movement and profile data, using assumptions about how an average body with your weight and age spends energy.

Every step up the ladder adds assumptions. That is why calorie error is larger than heart rate error in almost every validation study, and why it does not disappear even when the heart rate reading is good.

What did earlier studies of smartwatch calories find?

The 2026 study is new, but its main message is not. It fits a decade of research. (If you use an Apple Watch and want an Apple-specific walkthrough of earlier evidence, see our separate article on how accurate Apple Watch calorie tracking is.)

Stanford, 2017. Researchers tested seven wrist devices on 60 volunteers who sat, walked, ran and cycled while connected to heart monitoring and indirect calorimetry. Six of the seven devices had a median heart rate error under 5% during cycling. But no device estimated energy expenditure with an error below 20%. The authors concluded that most wrist devices "adequately measure HR in laboratory-based activities, but poorly estimate EE" (3).

Systematic review, 2020. A review of 158 publications covering nine wearable brands found that in lab settings Fitbit, Apple Watch and Samsung measured steps accurately and heart rate was more variable but often close. For energy expenditure its verdict was blunt: "no brand was accurate." In controlled settings, only 9.2% of energy expenditure comparisons were within 3% of the reference, and no brand stayed within 3% more than 13% of the time. Some brands mostly underestimated and others mostly overestimated (4).

Meta-analysis, 2020. A meta-analysis in the British Journal of Sports Medicine pooled 60 studies of wrist- and arm-worn monitors. It found that accuracy depended heavily on the type of activity, that results varied widely between studies, and that combining heart rate (or heat) sensing with movement data reduced error in most activity types (5).

Put together, the picture is consistent:

• Calorie estimates from wrist wearables are usually off by double-digit percentages.

• The direction of the error varies by brand, model and activity, so you cannot simply subtract a fixed amount.

• Heart rate helps the estimate, but it does not fix it.

• The 2026 study adds a new layer: body fat is one of the things that makes the error bigger.

Is Whoop more accurate for calories?

Whoop was not one of the watches tested in the 2026 study or in the 2017 Stanford study, so there is no head-to-head lab number for it in this evidence. No wrist-worn device can analyze the gases you breathe, so Whoop's calorie figure, like every other wearable's, is an estimate produced by an algorithm from sensor data. A strap without a screen does not change the physics: whatever device you wear, the calorie figure is a model's output.

If you use Whoop, the advice in this article holds: treat calories as a rough, relative number, and lean more on the signals closer to the sensor, such as resting heart rate and HRV trends.

Are treadmill calories accurate?

Gym machines estimate calories with a different, simpler set of inputs, and they have their own blind spots.

A treadmill knows its belt speed and incline. If you enter your weight, it can plug those into standard equations for the energy cost of walking or running. If you do not enter your weight, it uses a default. It does not know your heart rate (unless you hold the grips or pair a strap), your fitness, your body composition or how you are moving.

Two pieces of research show how that goes wrong:

• Leaning back on the handrails changes the cost of the workout, not the number. In a small study of 13 healthy adults walking on a 10% incline, leaning back while holding the handrails cut energy use to about 6.0 kcal per minute, compared with 8.8 kcal per minute walking at the same incline without support (6). The treadmill display has no way to know you are hanging on.

• Machines can overestimate. A study of an elliptical machine in 34 adults found it overestimated calories by about 100 calories per 30 minutes of moderate exercise compared with indirect calorimetry (7). That was one machine model, but it shows the same pattern as the watches: a formula that does not know your body.

If your treadmill and your watch disagree, neither is the "real" number. A practical way to use them:

1. Enter your current weight on the treadmill if it lets you.

2. Do not lean on the rails for incline walking unless you need them for balance; if you do need them, count the calories as lower than shown.

3. Use the same machine and the same settings when you compare sessions, and look at speed, incline, time and heart rate as your record of effort.

Which numbers on your watch can you trust more?

The researchers and earlier reviews point in the same direction. The FIU release passes on the authors' advice to rely more on heart rate and time in heart rate zones than on calories (2). Here is a simple way to sort what your watch shows:

More reliable

• Resting heart rate. Measured while you are still, which is when optical sensors work best. A resting heart rate that drifts up by several beats per minute over days often follows short sleep, illness, alcohol, heat or stress.

• HRV as a trend. A single reading is noisy, but your average over weeks, measured under similar conditions, is a meaningful signal of recovery and stress load. Our article on how accurate HRV is from different devices goes into the details.

• Sleep timing. When you fell asleep and woke up is generally easier for a wearable to pick up than the details of sleep stages.

• Heart rate during steady exercise. Cycling and steady walking or running tend to produce cleaner readings than lifting or boxing, where wrist movement and muscle tension interfere.

Less reliable

• Calories burned, active or total. Expect double-digit percentage error, and do not assume it cancels out.

• Calories during stop-and-start activity. Strength training, circuits, yoga and housework are harder for the model to read than steady cardio.

• Comparisons between brands. Two devices can disagree by a wide margin on the same workout. That says more about their formulas than about your body.

This is not about watches being "bad." It is about using each number for what it can do.

How to use calorie numbers without letting them run your plate

The study's authors leave room for one use: limited absolute accuracy "does not preclude potential utility for tracking relative or within-person changes over time," which they say still needs testing in real life (1). Kostrna's own advice to readers is to "track your own trends over time" (2).

Here is how to put that into practice:

1. Use calories as a relative signal. Comparing this week with last week on the same device can show whether you are moving more or less. Comparing 2,400 calories on one watch with 2,100 on another tells you nothing.

2. Keep the device and settings the same. Changing the workout type you select, wearing the watch on the other wrist, or swapping brands can shift the number without anything changing in your body.

3. Keep your profile current. Update your weight and age, because the estimate is built on them.

4. Do not "eat back" exercise calories one for one. The FIU release suggests treating calorie burn as a rough guide and rounding down. It also warns against overcorrecting by cutting food too hard, because your body still needs fuel to recover (2).

5. Check against outcomes, not the display. If your goal is weight, track your weight trend over two to four weeks alongside your intake. If your goal is fitness, track resting heart rate, how hard a familiar workout feels and how quickly your heart rate comes down afterward.

6. Notice when tracking stops helping. The same release says that if tracking becomes stressful, it is fine to stop fixating on the numbers.

The calorie number your watch shows at night also includes calories your body burns just to keep you alive. If you have ever wondered why the count keeps rising while you sleep, see how many calories you burn sleeping.

Living with a smartwatch: the questions people ask next

Should I eat back the calories my watch says I burned?

Not one for one. Most watches in the 2026 study overestimated calories during exercise, and earlier reviews found errors in both directions. If you are trying to manage weight, base food decisions on your weight trend over a few weeks rather than on the exercise number. If you are training hard and feel tired or keep getting sick, under-eating is a bigger risk than the watch being slightly off, and a doctor or registered dietitian can help.

Why do my Apple Watch and Garmin show different calories for the same workout?

Each company uses its own formula, built on its own training data, with different assumptions about your resting burn and about how heart rate and movement translate into energy. In the 2026 study, the same 58 people produced average overestimates ranging from about 3 kcal to about 69 kcal per session depending on the brand (1). Neither number is the true one. Pick one device and compare it with itself.

Does updating my weight make the estimate more accurate?

It helps, because weight is one of the inputs to the formula, and a wrong weight pushes every estimate in the wrong direction. But it does not fix the main limitation the 2026 study found. The watch still does not know your body fat or muscle mass, and error grew with body fat for every brand.

Are active calories more accurate than total calories?

Not necessarily. Active calories come from the exercise model; total calories add an estimate of your resting burn, usually calculated from your height, weight, age and sex. Both are estimates. In the 2017 Stanford study, no device got its energy expenditure estimate below 20% error (3).

Do steps count more accurately than calories?

Generally yes. The 2020 review found that Fitbit, Apple Watch and Samsung measured steps accurately in lab settings, while no brand was accurate for energy expenditure (4). Steps still have their own errors, especially pushing a stroller or carrying bags; see how fitness trackers count steps.

Is my watch's heart rate wrong too?

The 2026 study did not report heart rate accuracy at all, so it says nothing about it. Earlier lab studies found heart rate to be much closer to reference measurements than calories, especially during cycling and at rest (3, 4). A calorie error is not evidence that your heart rate reading is wrong.

Before you blame yourself

People who look up whether their watch's calories are accurate have often done everything "right." They closed their rings, logged their meals, ate back what the watch said they burned and still did not see the change they expected. The natural conclusion is "I must be doing something wrong" or "my body is broken."

The research points somewhere else. Calorie estimates in this study were typically 15% to 25% off and mostly too high. They were further off for people with more body fat. That is not a character flaw or a lack of discipline. It is a known limit of the tool.

So before you blame yourself, check what your body already recorded. Look at the numbers that sit closer to the sensor: your resting heart rate over the past few weeks, your HRV trend, how long and how regularly you slept. If your resting heart rate is creeping up and your sleep is short, the issue may be recovery, not effort. Change one thing, such as a consistent bedtime or one rest day a week, and watch the trend for one to two weeks.

When to see a doctor

A watch's calorie number cannot tell you why your weight is changing. See a doctor if you have:

• Weight loss or gain you cannot explain by changes in eating or activity.

• Ongoing tiredness or low energy that comes with a change in weight.

• A resting heart rate that stays noticeably higher than your usual for more than a week or two without an obvious reason such as illness or a new medicine.

• Thoughts about food, calories or exercise that feel out of control, or a pattern of restricting food or exercising to make up for eating.

A smartwatch is not the right tool for any of these. Bring the trends, not the calorie totals, to the appointment.

How to bring this up with your doctor

What to ask for, by name. "Could we look at why my weight is changing? Is there anything we should check or rule out?" If food or exercise feels hard to control: "I'd like to talk about my relationship with food and exercise."

What to bring. Two to four weeks of your weight trend, a rough food log, your typical activity, and your resting heart rate and sleep trends from your wearable. Leave the daily calorie totals as background, not evidence.

What to say, out loud. "My weight has changed by [amount] over [time] and I can't explain it. I've been tracking food and activity. My resting heart rate has gone from about [X] to [Y]."

If you're told to just eat less and move more. Ask: "I've been doing that and tracking it. Is there anything we should rule out first, and could you refer me to a registered dietitian?"

Start understanding your body

Learn what affects your energy, stress, sleep, and daily state. Get the app.

Explore your data

How Welltory helps — and what it cannot do

The limits first. Welltory is a general wellness product, not a medical device. It does not measure calories through gas exchange, it cannot make your watch's calorie estimate accurate, and it does not diagnose why your weight or energy is changing. Every calorie figure it shows is an estimate too, with the same kinds of limits described in this article.

What it can do is put the numbers closer to the sensor in front of you, so the calorie total stops being the headline.

It reads the watch you already have. Welltory receives heart rate, sleep, activity and workout data through Apple Health, Health Connect and Samsung Health, and connects directly to Garmin and Fitbit (data sources). Apple Watch, Samsung Watch and Google Pixel Watch (series 2, 3 and later) can be used for both activity and HRV; other watches and trackers usually provide activity data.

Effort measured against you, not an average. Welltory's activity analysis uses METs, a measure of how hard you are working compared with rest. Instead of standard average values, it personalizes them using your real-time heart rate, your resting heart rate and your Max METs, so the same walk can count as light for a fit person and moderate for someone else (how Welltory uses METs).

Active Calories that do not depend on starting a workout. In the Activity Report for Apple Watch users on iOS, which is still rolling out, Welltory calculates Active Calories itself rather than relying on whether a workout was logged (Activity Report). It is still an estimate, it was not tested in the studies described above, and we do not claim it is more accurate than your watch's number, so use it as a trend.

Your morning starting point. On the Today screen, available on iOS for Apple Watch and Oura users, Battery starts each morning from your resting heart rate on waking and the quality and quantity of your sleep. Health compares today's HRV with your own 60-day baseline (Today screen).

Context for the numbers. When Welltory marks a stress or rest episode on the Today screen (iOS with Apple Watch or Oura), it asks "What happened?" and you can answer in your own words: "leg day", "skipped lunch", "late dinner", "long walk". Those notes become tags, and over weeks My Patterns shows which ones tend to come with stress and which with recovery. Some My Patterns sections require a paid plan.

A record for your doctor. From the web app you can export your data as a CSV file (export guide) and bring resting heart rate, sleep and activity trends to an appointment. Patterns are associations to discuss, not a diagnosis.

How we made it

We based this article on the 2026 PLOS ONE study by Kostrna and colleagues at Florida International University, which we read in full, and on the university's press release. Error figures are taken from the paper unless we say they come from the release. We added the 2017 Stanford validation study, a 2020 systematic review of 158 publications, a 2020 meta-analysis of 60 studies, and smaller studies on treadmill handrail use and elliptical machines. Search questions came from what people in the United States most often ask about smartwatch, Fitbit, Garmin, Whoop and treadmill calorie accuracy. For the product section, we checked every function against the Welltory Help Center on October 22, 2026.

Made with AI tools, then edited and fact-checked by the Welltory team. See our Editorial & AI policy.

Jane Smorodnikova, co-founder of Welltory, built the methodology behind how we read physiological data.

Written by Tatsiana Yashyna.

Discounts for blog readers: up to 36% off

See what affects your energy, stress, sleep, and daily state with Welltory

Explore your data