Does exercise burn as many calories as you think?

The readout is unreliable, and the body claws back a quarter of the rest#

Does exercise burn calories? Yes — just far fewer than the number you were shown, and fewer still by the time the day is over. The question "how many calories does exercise burn" has two answers: the one on the console, and the one your body actually banks. They are not the same number, and the console's is not even reliably too high — it is simply unreliable. When Stanford researchers tested seven wrist-worn devices against clinical-grade reference measurement in 60 volunteers, not a single device achieved an energy-expenditure error below 20 percent — even though six of the seven tracked heart rate to within 5 percent during cycling¹. The sensor works. The calorie math on top of it does not.

Then the body gets a vote. Across the largest dataset of adult total and basal energy expenditure assembled (n = 1,754), energy compensation in a typical human averages 28 percent through reduced basal expenditure — only about 72 percent of the calories you burn through added activity survive as extra burn that day². So a "600-calorie" run is a number carrying at least a 20 percent error of unknown sign, and whatever the true figure was, roughly 72 percent of it survives the day. None of this makes exercise pointless — it makes exercise a poor accounting instrument and an excellent health one. This article traces exactly where the number leaks.

Where the exercise calorie leaks#

Four separate leaks sit between the effort you make and the deficit you keep. Each is independently documented, and they stack.

The first leak is a measurement problem you can partly route around. The other three are physiology, and they are the reason the treadmill number was never going to work as a budget line.

Leak one: the number on the screen is a guess#

Wrist devices and cardio machines do not measure calories. They measure something cheap — heart rate, motion, cadence — and run it through a model. The Stanford evaluation tested the Apple Watch, Basis Peak, Fitbit Surge, Microsoft Band, Mio Alpha 2, PulseOn and Samsung Gear S2 across 60 adults. Heart rate was largely fine. Energy expenditure was not: no device came in under 20 percent error, the Apple Watch was least bad and the Samsung Gear S2 worst, and error was systematically higher for men, for higher BMI, for darker skin tones, and during walking¹. The authors' own conclusion was to treat these numbers with caution in health programs.

A 20-percent-plus error on a 500-calorie readout is a 100-calorie-plus swing in a direction you cannot see — which is worse than being handed no number at all, since a figure that hides its error bar invites you to spend calories you may never have earned (the reason BurnWeek reports food as a range instead of a total). That swing is roughly the size of the daily deficit many people are trying to build. The same underlying problem — a plausible-looking number generated by a model rather than measured — is why calorie tracking apps have their own accuracy ceiling.

Leak two: your body balances the books behind your back#

This is the finding that reorganizes the whole topic. The intuitive model of energy expenditure is additive: baseline burn, plus whatever you do on top. The evidence says it isn't.

When researchers measured total energy expenditure by doubly labeled water in 332 adults across five populations — Ghana, South Africa, the Seychelles, Jamaica and the United States — while tracking movement with accelerometers, the relationship between physical activity and total expenditure did not rise in a straight line. It bent and flattened. Above roughly 219 counts per minute per day, each further increment of 100 counts per minute bought less than 50 additional calories a day of total expenditure, and the median difference in adjusted total expenditure between the least active and most active deciles was only 15 percent³. A change-point model beat a linear one. The authors' conclusion is blunt: public-health energy models underestimate activity's effect at low levels and overestimate it at high ones.

The mechanism appears to be exactly what compensation predicts — the body trims elsewhere. Careau and colleagues put a number on it: 28 percent on average, achieved through reduced basal energy expenditure, with the effect equal in men and women, young and old. Crucially, compensation was stronger in people carrying more body fat² — meaning the people most often told to exercise the weight off are the ones whose bodies push back hardest. That deserves to be said plainly rather than buried. This is the constrained energy expenditure model, and it is why exercise is the smallest and least dependable of the four components of your total daily energy expenditure.

The Hadza test: what happens when you really move#

The cleanest natural experiment comes from the Hadza, a hunter-gatherer population in Tanzania who forage on foot for their food. Researchers measured 30 Hadza adults with doubly labeled water. The men walked 11.4 ± 2.1 km a day, the women 5.8 ± 1.7 km — and their physical activity levels confirmed it, with Hadza men at a PAL of 2.26 ± 0.48 against 1.81 ± 0.21 for Western men⁴.

By the additive model, they should burn dramatically more calories per day than a desk worker. They don't.

Controlling for fat-free mass and age, Hadza foragers' daily energy expenditure was statistically indistinguishable from Westerners'. Lifestyle had no effect on total energy expenditure.

That is the study's own finding⁴: Hadza women's expenditure was similar to Western women's (n = 186), Hadza men's similar to Western men's (n = 53). Their conclusion follows directly — differences in obesity prevalence between populations "result primarily from differences in energy intake rather than expenditure." A physically demanding forager life does not buy a bigger calorie budget. It reallocates one.

Thirty people is a small sample, and cross-population comparisons carry confounds no amount of statistics fully removes — so the Hadza result cannot carry this argument by itself, and it isn't asked to. It works as one strand alongside the 332-person five-country accelerometry study and the 1,754-person compensation analysis: three designs with three different weaknesses, arriving at the same place.

What the actual weight-loss trials show#

Theory aside: when you put people on an exercise program and weigh them, what happens? A meta-analysis of 14 randomized controlled trials covering 1,847 participants found that isolated aerobic exercise produced a weight reduction of 1.6 kg at six months (95% CI −1.64 to −1.56) and 1.7 kg at twelve months (95% CI −2.29 to −1.11). The authors concluded that isolated aerobic exercise "is not an effective weight loss therapy" in overweight and obese patients⁵. A year of training, under 2 kg.

The individual-level picture is more interesting than the average. Thirty-five overweight, sedentary adults exercised five times a week for twelve weeks under supervision. The group mean loss was 3.7 ± 3.6 kg — close to what the exercise energy cost predicted, which looks like no compensation at all. But the individual range ran from −14.7 kg to +1.7 kg. Splitting the group by outcome, non-compensators lost 6.3 ± 3.2 kg while compensators lost 1.5 ± 2.5 kg⁶. Same prescription, same supervision, a four-fold difference — and the group average hid all of it.

The mechanism there is largely behavioral: appetite and intake move in response to training, in some people much more than others. That's covered in exercise compensation and eating back your calories.

So what actually moves the deficit?#

Intake, logged consistently. Not because exercise is worthless, but because you can't outrun a bad diet: the intake side has no readout error of 20 percent baked into the hardware, no 28 percent metabolic clawback, and no plateau above a threshold. When you eat 200 fewer calories, you have eaten 200 fewer calories. When you burn 200 more, you have maybe kept 145 of them — 200 less Careau's 28 percent — and you may quietly eat some back.

Three practical conclusions follow, and none of them is "stop training":

1. Stop counting exercise calories toward your budget. Don't eat them back. If the readout is 20-plus percent wrong in an unknown direction and the body reclaims roughly a quarter of whatever was real, the defensible accounting entry is zero. Set your deficit from intake and let training be a bonus you never spend.
2. Chase NEAT, not the console. The flattening in the Pontzer data occurred at the top of the activity range; the same work found that at low-to-moderate levels, activity's effect is underestimated. Getting a sedentary person moving is where the energetic return actually lives — see what NEAT is.
3. Train for what training reliably delivers. Cardiovascular health, strength, mood, and the fat-free mass that sets the resting burn dominating your total — the biggest term in metabolism explained, and the one exercise genuinely protects. Those returns aren't compensated away.

Exercise is not a calorie lever with a bad exchange rate. It is a superb intervention being measured with the wrong instrument, on the wrong outcome.

FAQ#

How many calories does exercise actually burn?#

Fewer than your device says, and fewer than that once the day nets out. Wrist-worn trackers showed energy-expenditure errors above 20 percent for every device tested, and a typical person compensates about 28 percent of activity calories through reduced basal expenditure. A useful rule: treat the readout as an upper bound and don't budget against it.

If exercise barely helps weight loss, should I stop?#

No. The evidence says exercise is a weak calorie lever, not a weak intervention. Aerobic exercise alone produced only about 1.7 kg of weight loss at twelve months across 14 trials — but it delivers cardiovascular, strength, and mood benefits, and it protects the fat-free mass that sets your resting metabolic rate. Judge it by those outcomes, not by the console.

Why do some people lose weight from exercise and others don't?#

Mostly compensation, and it varies enormously between individuals. In one twelve-week supervised program the group average looked fine while individual results ran from 14.7 kg lost to 1.7 kg gained; sorting by outcome, compensators lost 1.5 kg and non-compensators 6.3 kg. Appetite response and unconscious reductions in other daily movement explain most of that spread.

Sources#

1. Shcherbina A, et al. Accuracy in wrist-worn, sensor-based measurements of heart rate and energy expenditure in a diverse cohort. J Pers Med. 2017;7(2):3.
2. Careau V, et al. Energy compensation and adiposity in humans. Curr Biol. 2021;31(20):4659-4666.
3. Pontzer H, et al. Constrained total energy expenditure and metabolic adaptation to physical activity in adult humans. Curr Biol. 2016;26(3):410-417.
4. Pontzer H, et al. Hunter-gatherer energetics and human obesity. PLoS One. 2012;7(7):e40503.
5. Thorogood A, et al. Isolated aerobic exercise and weight loss: a systematic review and meta-analysis of randomized controlled trials. Am J Med. 2011;124(8):747-755.
6. King NA, et al. Individual variability following 12 weeks of supervised exercise: identification and characterization of compensation for exercise-induced weight loss. Int J Obes. 2008;32(1):177-184.