Nearly 4 in 10 US adults — roughly 100 million people, and about 560 million worldwide — now wear a smartwatch or fitness tracker, leaning on it to answer one of the most basic questions in health and weight management: how many calories did I actually burn? A new peer-reviewed study out of Florida International University (FIU) suggests the answer that number gives you may be substantially wrong — and wrong in a way that hits hardest exactly the people who rely on it most.

The study, in plain terms

Researchers at FIU’s Medical Photonics Laboratory, led by exercise scientist Jason Kostrna, put 58 Hispanic adults aged 18 to 50 — deliberately recruited to span a range of body sizes and skin tones, from medium to deep — through a controlled cycling session on a recumbent bike. Each participant wore four popular smartwatches at once — the Apple Watch Series 8, Fitbit Sense 2, Samsung Galaxy Watch 5 and Garmin Forerunner 955 — while breathing into a clinical-grade metabolic analyzer that measures actual energy expenditure with medical precision. That device served as the “gold standard” the smartwatches were checked against. The findings were published in the journal PLOS One.

The topline result: every smartwatch missed the mark, with typical errors landing around 15% to 25%, and the worst-performing devices drifting far higher — in some cases 50% to 100% or more off the true number. Put concretely: if the lab equipment measured a true burn of 100 calories during a session, some watches displayed figures as high as 150 to 200 calories for the same effort.

The part that should worry health-conscious users most

The headline number is bad enough on its own, but the more consequential finding is what drove the error: the higher a participant’s body fat percentage, the less accurate every single watch became — regardless of brand. This wasn’t isolated to one device; it held across all four trackers tested, though the size of the effect varied significantly by brand. Notably, the researchers found skin tone did not have a significant effect on the error under the conditions they tested, isolating body composition as the specific variable driving the inaccuracy.

That pattern is a genuine problem, not a footnote. People carrying more body fat are often precisely the population using calorie-burn data to manage weight, plan meals and stay motivated — and, as a group, they already face higher risk for cardiovascular and metabolic disease. If the device meant to guide their decisions is least reliable for their bodies specifically, it risks quietly sending exactly the people who need accurate data the furthest in the wrong direction, without any indication on the watch face that something is off.

Brand-by-brand, results weren’t equal

The study found real differences between devices. Apple’s watch came out as the most accurate of the four. Garmin and Samsung overestimated calorie burn the most. Fitbit’s results were murkier still — the device occasionally reported implausible readings, including as little as one calorie for an entire workout session, and sometimes failed to produce a reading at all; so much of its data was unreliable that researchers had to exclude a large share of it from the final analysis rather than draw firm conclusions about its accuracy.

Kostrna and his team don’t yet have a definitive explanation for why body fat specifically degrades accuracy, though he suspects it traces back to the proprietary algorithms each company uses to convert raw sensor data — heart rate from light-based optical sensors, plus movement from accelerometers — into a calorie estimate. As Kostrna put it, nobody outside these companies knows what population the training data behind those algorithms actually represents, which is precisely why independent studies testing real, diverse bodies matter.

This isn’t a one-off finding

The FIU study lands on top of a body of prior research raising similar doubts. A widely cited Stanford University study years earlier tested seven wrist-worn trackers, including early Apple Watch, Samsung and Fitbit models, and found all of them “way off the mark” on calorie estimates, with the least accurate device overstating energy burned by as much as double. Other academic reviews have documented heart-rate tracking errors that grow substantially during exercise itself, when repetitive motion can be mistaken by a device’s light-based sensor for a heartbeat — a distortion that compounds directly into the calorie math, since energy expenditure estimates lean heavily on heart-rate data.

Taken together, the pattern across nearly a decade of independent research is consistent: consumer wearables are reasonably good at counting steps and tracking basic movement, but calorie-burn estimates — arguably the single number most people actually check to make decisions — remain some of the least trustworthy data these devices produce.

So should you stop trusting your watch?

Not entirely — but the researchers are direct about what these numbers are, and aren’t. Kostrna’s own framing is blunt: treat the calorie figure your watch shows as a rough ceiling, not a precise measurement, since you could plausibly be off by hundreds of calories over a week and end up in a calorie surplus while believing you’re in a deficit. For anyone using a smartwatch to manage weight, plan meals around a calorie target, or make clinical decisions based on tracked activity, that’s a meaningful caveat rather than a minor technical footnote — and it argues for treating the number as one directional signal among several (how you feel, how your weight trends over weeks, other tracked metrics) rather than as ground truth to be logged and acted on precisely.

For the industry, the fix the researchers point to is straightforward in principle, if not necessarily easy in practice: train and validate these estimation algorithms on genuinely diverse body types and compositions, rather than whatever convenience sample of test subjects a company happened to use during development — and be transparent enough about that testing for outside researchers, and eventually regulators, to check the claims.