Your Smartwatch Knows Your Pulse. It Is Guessing About Your Calories.

New validation studies show a sharp divide in what wrist wearables measure well. Knowing the difference changes how you should use them.

By Qais S. Dirar, MBBS, MPH, DrPH

At the end of a workout, millions of people glance at their wrist and accept a number: 487 calories burned. It feels like a measurement. In most cases it is an estimate, built from a few sensor readings and an algorithm. The device's confident display hides a wide margin of error. As someone who teaches people how to judge measurement tools, I find the gap between how precise these numbers look and how precise they are worth examining.

The latest validation research tells a consistent story. Wrist wearables have become good at one thing and remain unreliable at another.

What they get right: heart rate

A validation study presented at the World Physiotherapy Congress in 2025 asked 30 recreationally active adults to wear an Apple Watch for several days and then complete a maximal treadmill test. Heart rate from the watch was compared with a Polar H10 chest strap, a widely used reference. Agreement was close: the average difference was 1.1 beats per minute (World Physiotherapy Congress, 2025).

That matches a broader pattern. Optical heart-rate sensing, which reads blood-volume changes through the skin, is now quite accurate for healthy people during steady exercise. If you use your watch to keep runs in a target zone or to notice a trend in your resting heart rate, you are on reasonably firm ground.

What they get wrong: calories and fitness

The same study tells a different story for energy expenditure and VO₂max, the standard measure of cardiorespiratory fitness. Compared with indirect calorimetry, which measures the oxygen you consume and the carbon dioxide you exhale, the watch underestimated energy expenditure. The limits of agreement were wide, running from about 53 kilocalories too low to 68 too high for individual readings. It also underestimated VO₂max by an average of 5.5 ml/kg/min, a mean absolute percentage error of about 13%.

Those limits of agreement matter more than the average. A small average bias can hide large errors for any one person on any one day, and you are always one person on one day.

A separate 2025 study in Physiological Measurement tested an Apple Watch Series 7 against spirometric calorimetry while 35 healthy young adults cycled at increasing intensity. The watch tracked the direction of change well, with a correlation of 0.93. It still underestimated energy expenditure by an average of 27.4 kilocalories, with limits of agreement of about 62 kilocalories (Physiological Measurement, 2025).

That is the essential distinction. A wearable can rank your sessions reliably, telling you that today's ride was harder than yesterday's, while still being off on the absolute number.

Accuracy is not universal

Most validation studies recruit healthy, young, fit volunteers, the people for whom the algorithms were tuned. A 2025 study in JMIR Cardio looked further. It tested a medically certified wrist band in 57 people: those with heart failure, those with stable coronary artery disease, and recreational athletes, all performing everyday activities (JMIR Cardio, 2025).

In the athletes, heart-rate accuracy was good. In the patients, the device significantly underestimated heart rate, with poor reliability in heart failure and only fair reliability in coronary disease. Energy-expenditure reliability was poor or fair across groups. The authors concluded that algorithms need to account for disease and medication before such devices are relied on in patient care.

For readers, the lesson is straightforward. Published accuracy figures describe the people who were studied. If your physiology, medications or activities differ from theirs, the numbers may not transfer to you.

Why calories are so hard

Heart rate is a direct physical signal. Energy expenditure is not measured at all. It is inferred from heart rate, movement and the age, sex, height and weight you entered into the app. Two people with the same heart rate can burn very different amounts of energy, depending on fitness, body composition, the type of exercise and even heat. An algorithm can only guess at those differences from the outside.

How to use the numbers wisely

None of this makes wearables useless. Used well, they are powerful tools for building habits. A few principles help.

Trust trends over single readings. Week-to-week changes in your resting heart rate, step counts or training load are more meaningful than any individual figure.

Don't eat back your calories. Basing food intake on a watch's calorie estimate imports its error straight into your diet, which may explain some stalled weight-loss efforts.

Treat fitness estimates as rough ranges. A VO₂max reading can motivate you, but it is not a clinical test. If precision matters, for example in planning cardiac rehabilitation, use a laboratory assessment.

Use clinically validated devices for medical decisions, and talk to a clinician before changing treatment based on wearable data.

The bigger picture

Wearables are moving quickly from lifestyle gadgets into clinical care. That transition depends on evidence like this. These studies are not an indictment of the technology. They describe its limits, which is exactly what responsible adoption requires. The heart-rate results show how far the field has come. The calorie results show how much of what we read on our wrists is still an educated guess.

This article summarises published research and is not medical advice.

---

Qais S. Dirar, MBBS, MPH, DrPH, is a physician and epidemiologist who writes about metabolic health, clinical evidence and health systems.