Wearables occupy an odd space in consumer technology: they’re marketed with the confidence of a medical device and worn with the casualness of a bracelet, and the actual research sits somewhere in between those two framings — genuinely useful for some things, meaningfully unreliable for others, and rarely as precise as the clean number on the screen suggests.
What they’re actually accurate at
Heart rate is the strongest category. A systematic review of commercially available wearables’ accuracy for measuring steps, energy expenditure, and heart rate found that these devices are generally most reliable at heart rate tracking relative to the other two metrics. A separate validation study specifically testing optical heart rate sensors — the light-based sensors on the back of most wrist-worn devices — found reasonable accuracy under steady conditions across both younger and older adults, which is the use case most people actually rely on: a resting or steady-state heart rate reading during the day.
Where accuracy consistently breaks down is during what researchers call transient states — the moments right after you start moving, or when your heart rate is changing quickly. A more recent validation study comparing several popular wrist-worn devices found performance declined across every device tested during those transitions, with motion onset and rapid heart rate changes producing the largest errors. Some devices handled this better than others in that particular comparison, but the review’s broader conclusion was blunter: there’s no single brand or model that performs well across every measure, and nothing in the current generation of consumer wearables qualifies as a genuine “gold standard.”
Where the numbers get shakier
Step counting looks precise — a specific number, updated all day — but the underlying accuracy varies more than that precision suggests. Research has found devices from Apple and Samsung showing the strongest validity for step counts specifically, with Apple, Fitbit, and Garmin devices landing within an acceptable accuracy range only around half the time across the studies reviewed. Age and sex have also been shown to affect how accurately a given device counts steps, meaning the same device can perform differently well depending on who’s wearing it.
Energy expenditure — the calorie-burn estimate most trackers proudly display — is the weakest category by a clear margin. The same systematic review found that no brand tested fell within an acceptable accuracy range for energy expenditure at all. That’s worth sitting with for a moment, because “calories burned” is often the number people treat as most actionable, whether they’re deciding how much to eat or how hard to push a workout, and it’s the one the research trusts least.
The one area with genuinely strong clinical evidence
There’s a notable exception to the “no gold standard, mixed accuracy” pattern above, and it’s worth separating out because it gets conflated with the rest: heart rhythm irregularity detection, specifically for atrial fibrillation. This isn’t a marginal claim. A systematic review and meta-analysis published in the peer-reviewed journal JACC: Advances found pooled sensitivity of 94.8% and specificity of 95% for Apple Watch-based detection of atrial fibrillation against clinical-grade comparison. That’s a meaningfully different evidence tier than step counts or calorie estimates — closer to genuine diagnostic performance than casual fitness tracking.
The regulatory status reflects that distinction too: the feature has FDA clearance through the agency’s 510(k) pathway, specifically for people with no prior AFib diagnosis, as a notification prompting further evaluation rather than a standalone diagnosis. That’s an important nuance the “FDA cleared!” marketing sometimes glosses over — it’s cleared to flag a possible irregularity worth checking, not cleared to replace a cardiologist. But within that narrower claim, the accuracy data is genuinely strong, and it’s the clearest example of a wearable feature earning trust that goes beyond “reasonable estimate.”
Does wearing one actually change anything?
Accuracy is one question. Whether the data changes behavior is a separate one, and it’s less settled. A tracker can be reasonably accurate at what it measures and still not meaningfully change what someone does day to day — or it can be somewhat inaccurate and still be genuinely useful as a rough trend indicator, the same way a scale doesn’t need to be accurate to the gram to usefully show you’re trending up or down over a month.
That reframing — trend over precision — is closer to how the more credible parts of the research treat wearable data. A single day’s step count or heart rate reading carries real measurement error. The shape of that number over weeks or months is more informative than any single reading, and it’s a more honest way to use a device whose moment-to-moment numbers don’t hold up to close scrutiny.
The practical takeaway
Trust a wearable’s heart rate readings during calm, steady periods more than during a sudden sprint or an intense interval set. Treat the step count as a reasonable approximation rather than an exact figure, particularly if you’re using a brand outside the higher-accuracy group. And treat the calorie-burn number with real skepticism — it’s the metric with the weakest evidence behind it of the three most commonly tracked, in every major review of this technology so far.
None of that makes wearables useless. It makes them closer to what a good telehealth visit already is: a genuinely useful tool for the right kind of monitoring, with real, specific limits that are worth knowing rather than assuming away. And like most everyday AI tools, the device is most useful when a person treats its output as a helpful estimate to work from rather than a verdict to trust without question — a habit that also happens to matter for getting decent sleep, given how many people now check a sleep score before they’ve even gotten out of bed.
