Fifty-six studies. Three metrics. One of them is 28 percent wrong — and it’s the one you use to decide what to eat.
Ninety-seven percent of studies say your Apple Watch heart rate is accurate. Sixty-two percent say your calorie count is not.
Your Apple Watch says you burned 600 calories. A 56-study meta-analysis says the real number is closer to 432.
That is 168 calories your watch invented. Not estimated generously. Not rounded up. Invented. They never existed as heat leaving your body, as muscles contracting, as energy spent. The watch's algorithm produced them, and you may have eaten them back for dinner.
Researchers at the University of Mississippi analyzed every Apple Watch accuracy study published through November 2024 — 56 studies, 270 separate measurements, thousands of participants — and found the device overestimates calorie burn by an average of 27.96 percent. Nearly three times the 10 percent threshold the International Electrotechnical Commission considers valid for a measurement device.
The study was published in April 2025 in Physiological Measurement and preregistered with PROSPERO before a single database search was run. Most participants were healthy adults under 40 — the exact demographic most likely to check their wrist after a gym session and plan dinner around the number.
Apple Watch calorie tracking overestimates by 28 percent on average. No brand, model, or activity type passes the accuracy threshold. The correction: eat back 70 to 75 percent of displayed exercise calories.
- Seven generations of Apple Watch improved heart rate accuracy nearly four times over. Calorie accuracy barely moved — 31.5 percent error on Series 1, 26.2 percent on Series 6. The problem is not old hardware.
- A separate review tested 29 brands and 72 devices. Every single brand exceeded 30 percent calorie error. Switching to Fitbit, Garmin, or Samsung does not fix this.
- The practical correction: eat back roughly 70 to 75 percent of the exercise calories your watch displays. The remaining 28 percent is algorithm inflation that 56 studies quantified.
- Calorie accuracy varies by what you are doing. Rest is the worst at 43.3 percent error. Running is the least inaccurate exercise category at 21.6 percent. Every activity type still fails the 10 percent validity threshold.
Three Grades From One Device
Here is where the study gets strange.
The same Apple Watch, on the same wrist, during the same workout, earns three completely different grades. Heart rate: 4.43 percent error. Thirty-six out of 37 studies — 97.3 percent — confirmed the measurement is accurate.
Steps: 8.17 percent error. Eleven out of 13 studies — 84.6 percent — said the count is reliable.
Calories: 27.96 percent error. Eight out of 13 studies — 61.5 percent — concluded the measurement is not valid.
Three metrics. One device. One wrist. An A+ on heart rate, a solid B on steps, and an F on the only number most people use to decide what to eat.
The heart rate sensor works by flashing green light into your skin and measuring how the reflected light changes as blood pulses through your wrist. That measurement has gotten remarkably precise. The calorie estimate depends on that same heart rate data — plus motion data, plus a proprietary algorithm that converts all of it into energy expenditure. The algorithm is where the translation breaks.
Your watch reads your pulse correctly and then miscalculates what that pulse cost you.
Seven Generations, One Unsolved Problem
Apple Watch Series 1 measured heart rate with 5.8 percent error. By Series 7, that had dropped to 1.5 percent. Nearly four times more accurate, generation after generation, each one meaningfully better than the last.
Calorie accuracy across the same span: Series 1 at 31.5 percent error. Series 6 at 26.2 percent. Seven generations of hardware and software engineering. The calorie gap barely closed.
The researchers noted this contrast directly. For heart rate and steps, they observed the error range narrowing with newer models. For calories, no such improvement appeared. Apple's engineers solved the heart rate problem. The calorie problem — converting what your body is doing into how much energy it spends — appears to be fundamentally harder.
And the error is not evenly distributed. If you wear your watch while sitting at your desk, the calorie estimate is off by 43.3 percent. That is the highest error rate in any activity category the researchers measured. The number your watch shows at rest — the baseline your activity rings calculate from — is the least accurate number on the entire screen.
Light-intensity activity came in at 28.9 percent error. Running: 21.6. Resistance training: 20.0. Free-living daily activity: 18.2. Every single category failed the 10 percent validity threshold. But the baseline — the resting number — was the worst of all.
The Escape Hatch That Does Not Exist
The natural response to learning your Apple Watch is 28 percent wrong on calories is to consider switching brands.
A separate systematic review tested that instinct. Germini and colleagues reviewed 57 studies covering 72 devices from 29 different brands — Fitbit, Garmin, Samsung, Polar, Whoop, and two dozen others. For energy expenditure, the error rate was above 30 percent for every brand tested [1].
Every single one. Not most. All.
The wearable technology market ships hundreds of millions of devices per year. None of them have solved calorie estimation. Heart rate and steps are largely accurate across brands, just as they are on the Apple Watch. Calories remain the unsolved problem industry-wide.
Switching devices is rearranging your wrist, not fixing the measurement.
Midnight Pacing for a Number That Was 28 Percent Wrong
If the error were just academic, it would be interesting and harmless. It is not academic.
A Fortune investigation in January 2025 documented what the calorie number does to real people [2]. Ryann Nicole described pacing her living room at midnight to close her Apple Watch rings. Jessica Post shook her wrist during class to register movement hours. People wrote on Reddit about walking at 11:30 PM feeling "ashamed" that their rings were not complete.
The language people use about their relationship with their watch is telling: "bullied," "consumed," "controlled."
A 2023 study found that when Apple Watch step counts were experimentally manipulated to show deflated numbers, participants experienced reduced self-esteem and ate more unhealthily. The number on the screen did not just inform behavior. It altered how people felt about themselves.
These people reorganized their evenings, their rest days, and their relationship with food around a calorie number. That number was 28 percent wrong.
They tested twenty-nine brands. Seventy-two devices. Every single one failed at calories.
The Bowling Ball You Never Picked Up
One hundred sixty-eight phantom calories per workout. The number sounds abstract until you multiply it.
168 calories a day, seven days a week: 1,176 phantom calories every week. Over a year: 61,320 calories that never existed as actual energy expenditure. At roughly 3,500 calories per pound of body fat, that is 17.5 pounds.
A bowling ball. Gained entirely from eating back a number your watch made up.
The MyFitnessPal community has debated for years whether to eat back exercise calories. Thousands of forum threads. The same argument on repeat. MyFitnessPal's own official guidance: eat back only 50 to 75 percent of your exercise calories.
The meta-analysis quantified the inflation at 28 percent. Eating back 72 percent of your exercise calories — 100 minus 28 — lands in the middle of MyFitnessPal's recommended range. The workaround the community argued about for years turns out to match the correction factor 56 studies confirm.
What the Researchers Actually Said
Before the correction factor becomes a new rule, the study deserves the honesty it earned.
The researchers acknowledged genuine progress. For heart rate, the data shows clear generational improvement — Series 1 at 5.8 percent error to Series 7 at 1.5 percent. That is real engineering. And the overall calorie error of 27.96 percent is an average across all studies, not a single precise number, which means individual study accuracy varies considerably.
No researcher the study cited defended 27.96 percent as acceptable for calorie tracking. The most generous framing from the scientific community: the device is excellent for tracking habits and staying motivated, but the calorie number should not be treated as precise truth.
The population was 80.4 percent healthy and 73.2 percent under 40. If you are older or have a clinical condition, these specific numbers may not map to your body the same way. The study covers Apple Watch specifically — other brands may differ in individual performance even though the Germini review found the category-wide pattern holds.
One finding adds unexpected credibility. The researchers tested for publication bias — the tendency for journals to publish only positive results — and found none for the calorie measurement. Bias was detected for heart rate and step data, but not for calories. The most damning number in the study is also the most statistically robust.
A Report Card, Not a Breakup Letter
The meta-analysis is a report card. Not a verdict to stop using the device.
Heart rate: trust it. 4.43 percent error, improving every generation, validated by 97.3 percent of studies. If your watch says your heart rate hit 155 during a sprint interval, that number is almost certainly close.
Steps: trust it, mostly. 8.17 percent error overall, though accuracy drops for adults over 40 (10.9 percent) and on treadmills (10.1 percent). If your watch says you took 8,000 steps, the real number is likely in range.
Calories: treat it as a rough direction, not a precise number. The 27.96 percent average error means a 600-calorie workout display corresponds to roughly 430 actual calories. The practical fix is straightforward: do not eat back 100 percent of exercise calories. The data suggests eating back roughly 70 to 75 percent puts you closer to the actual expenditure.
The watch is still useful. The calorie metric is still directionally informative — a harder workout will still show more calories than an easier one, and that relative signal is worth having. What changes is your trust in the precision of one specific number. The one that happens to be the number most people base their food decisions on.
You now know something most people who check their wrist after a workout do not: which numbers to believe and which one to adjust. The watch did not lie. The algorithm just is not there yet.
Which raises the next question. If the device that tracks your exercise calories is 28 percent wrong, how accurate is the equation your calorie calculator uses to estimate what your body needs in the first place?
Your watch is still a useful comparison tool. A harder workout shows more calories than an easier one, and that relative signal is real. What changes is how you use the absolute number.
For daily calorie planning, the meta-analysis puts the correction at roughly 72 percent of displayed exercise calories. A 500-calorie workout display corresponds to about 360 actual calories burned. A 300-calorie walk corresponds to about 216.
If you are in an active calorie deficit and eating back exercise calories at face value, the 28 percent inflation may be erasing roughly a third of your deficit every session. The watch is not lying about effort — it is inflating the currency you are spending at dinner.
What other research found
What this means for you
Every session where you eat back your full exercise calorie display, the 28 percent inflation eats into your deficit. A 500-calorie daily deficit shrinks to roughly 330 when you add back a 600-calorie workout at face value — because 168 of those calories never existed as energy your body spent.
The error is not random noise. The mean bias is consistently positive (+0.30 kcal/min), which means the watch consistently works against anyone using it to create a calorie gap. The correction is the same math the meta-analysis quantified: treat roughly 70 to 75 percent of the displayed exercise number as real.
Heart rate accuracy holds steady across age — 4.4 percent error under 40, 4.6 percent over 40. That green metric stays green. But step count accuracy drops. Under 40: 4.3 percent error. Over 40: 10.9 percent — crossing the validity threshold that separates reliable from unreliable.
Add treadmill walking (10.1 percent error) and light-intensity activity (23.9 percent), and the over-40 step count becomes unreliable in the conditions where many older adults actually use it. Your personal report card just lost a second metric. Heart rate: still trustworthy. Steps: now borderline. Calories: still the same 28 percent error everyone gets.
The calorie error changes with the activity. Running: 21.6 percent error — the closest any exercise category gets to the threshold, still more than double it. Resistance training: 20.0 percent. Cycling: 27.2 percent. Treadmill: 27.6 percent.
In practical numbers: your 400-calorie run display is closer to 314 actual calories. Your 500-calorie cycling session is closer to 364. Every activity type still fails the threshold — but running and lifting are the least inflated. If you train multiple activities, the correction varies by session type.
The Move ring calculates your daily active calorie goal from a resting metabolic estimate — the baseline number your watch produces when you are sitting still. That resting estimate has the highest error of any category the researchers measured: 43.3 percent.
The goal your ring is chasing was set by the least accurate number on your screen. Some days the ring feels impossible to close without an evening walk. Some of that gap between your effort and the ring's demand is real. Some of it is a 43 percent resting error setting a target that never quite matched your actual metabolism.
Before you change anything
The participants across these 56 studies were 80.4 percent healthy and 73.2 percent under 40. That is a specific demographic — largely young, healthy adults in controlled laboratory settings. If you have a metabolic condition, a cardiac issue, or you are significantly older than this study population, the 28 percent figure is directional but may not be your exact number. The study tested Apple Watch specifically. Other brands show similar category-wide failure in the Germini review, but your individual device's accuracy may differ from the brand average.
The 27.96 percent is an average across 34 studies that reported percent error — not a single precise measurement. Individual study accuracy ranged widely. One person's watch might be 15 percent off. Another's might be 40 percent off. The 28 percent is the center of that spread — not a promise about your specific device. The equipment used to measure the "true" answer also varied across the 56 studies — different heart rate monitors, different motion sensors, different lab setups — which may have introduced some inconsistency. And the algorithm itself is a black box: the study can measure how wrong the number is, but not where inside Apple's software the error originates.
Not all parts of this study carry the same weight. 'Wearable calorie tracking is inaccurate' — high confidence. Fifty-six studies all pointing the same direction, backed by a separate 57-study review covering 29 brands. That is as close to settled as exercise science gets. 'The average error is about 28 percent' — moderate confidence. That number is a group average across very different studies. Some found 15 percent error, others found 40. The truth for any single workout depends on what you were doing and which watch you wore. 'Your specific watch on your specific workout is 28 percent wrong' — low confidence. Individual sessions swing wildly. The watch might nail one run and miss the next by half. The 28 percent is what happens across thousands of measurements — not a guarantee for Tuesday's leg day.
If the device tracking your exercise calories is 28 percent wrong, what about the equation estimating how many calories you need in the first place?
The Mifflin-St Jeor equation powers most calorie calculators, including MyFitnessPal. It explains only 71 percent of the variance in resting metabolic rate. Two numbers feeding into the same meal plan. Both carrying double-digit uncertainty.
If exercise calories are inflated and baseline calories are imprecise, how accurate is your food tracking?
Research on dietary self-reporting has its own uncomfortable answer. The calorie equation behind the app, the exercise number on the watch, and the food log on the screen — all three carry error. The question is how much.
What This Study Found
All findings from this paper, in plain language.
- Apple Watch calorie tracking fails the accuracy threshold in every subgroup the researchers tested — 27.96 percent average error, nearly three times the 10 percent limit scientists consider valid for a measurement device.
- Apple Watch heart rate tracking passes — 4.43 percent average error, well below the threshold, with 97.3 percent of published studies confirming accuracy.
- Step counting is accurate overall at 8.17 percent error, but breaks down for adults over 40 (10.9 percent), on treadmills (10.1 percent), and during light activity (23.9 percent).
- Resting calorie estimation is the worst at 43.3 percent error. Free-living daily activity is the best at 18.2 percent. Running (21.6 percent) and resistance training (20.0 percent) fall in between. Every activity type exceeds the threshold.
- Step accuracy drops with age. Under 40: 4.3 percent error. Over 40: 10.9 percent — crossing the line between reliable and unreliable.
- Heart rate accuracy improved nearly four times across Apple Watch generations (5.8 percent on Series 1 to 1.5 percent on Series 7). Calorie accuracy barely moved: 31.5 percent on Series 1, 26.2 percent on Series 6.
- On average, the watch slightly overestimates calories (+0.30 kcal/min), slightly underestimates heart rate (−0.12 bpm), and slightly undercounts steps (−1.83 steps/min). But wide Limits of Agreement mean individual measurements can swing far in either direction.
- Among published studies, 97.3 percent confirmed heart rate accuracy while 61.5 percent concluded calorie tracking is not valid. Same device, opposite conclusions depending on which metric you ask about.
- No publication bias was detected for the calorie finding (p = 0.17). Bias was found for heart rate and step data. The most damaging finding is the one with the cleanest statistical foundation.
- Removing low-quality studies from the analysis did not change the results. Calorie error remained at 27.71 percent — virtually identical to the full 27.96 percent.
- The small average bias (+0.30 kcal/min, about 6 percent) and the large percent error (28 percent) are not contradictory. Mean bias allows overestimates and underestimates to cancel. Percent error captures total magnitude regardless of direction.
- No single accuracy number captures every condition. Step count error ranged from 1.2 percent for walking to 23.9 percent for light activity. Age, activity type, intensity, and device series all shift the accuracy picture.
- When the reference device uses the same measurement method (accelerometry) as the Apple Watch, step count agreement improves. Shared technology may explain some agreement beyond true accuracy.