Smartwatches: What They Measure and Estimate

Direct, estimated and calculated

Studies name exact models. This page does not rank brands, and a result for one watch should not be assumed for another or for a newer version.

Heart rate and exercise

A review of 65 articles found the Apple Watch had a mean absolute percentage error under 10% for heart rate in 2 studies, but over 30% for energy expenditure for every brand. [1] A meta-analysis of 56 Apple Watch studies found a mean heart-rate bias of -0.12 beats per minute with wide limits of agreement (about -11 to +11), and no heart-rate subgroup above a 10% error threshold. [2] In a treadmill study of the Apple Watch Series 6 and 7 against a 12-lead ECG, validity was good, but error grew as running speed increased. [3] Evidence label: promising but limited, for the models studied.

Irregular-rhythm detection

A 2025 meta-analysis of 26 studies (17,349 patients) found pooled sensitivity of 95% and specificity of 97% for atrial fibrillation detection by smartwatches. [4] Results varied by device: 94% sensitivity for Apple Watch, 97% for Samsung devices and 89% for the Withings ScanWatch, with very high heterogeneity between studies. [4] A separate review of wearable trackers found pooled sensitivity of about 94% and specificity of about 95% for atrial fibrillation, and also about 82% sensitivity but lower and more uncertain specificity (about 63%) for fall detection. [5] Pooled numbers hide large differences between studies, and a figure for one model or population should not be applied to everyone. An alert is a reason to see a clinician, not a diagnosis. Regulatory status of any specific watch feature: not verified here; check the FDA or your country's regulator for the exact model and function.

Sleep

In a 35-person overnight lab study, the Apple Watch Series 8 detected sleep versus wake with sensitivity of 95% or more, but underestimated deep sleep by 43 minutes and overestimated light sleep by 45 minutes against polysomnography (the sleep-lab standard). [6] Across 11 consumer sleep trackers, agreement with polysomnography varied substantially. [7] See Wearables vs Sleep Studies.

Skin tone and fit

A systematic review of 10 studies found that four reported lower heart-rate accuracy in darker-skinned people, four found no effect and two were mixed; the authors called the evidence inconclusive. [8] Meta-analysis of wearable pulse rate studies found no significant bias by pigmentation, but wide limits of agreement, wider in the darker pigmentation group. [9]

Things a smartwatch cannot do

The FDA says not to buy or use smartwatches or smart rings that claim to measure blood glucose, and it has not authorized any that measure or estimate glucose on its own. [10] Inaccurate glucose readings can lead to wrong diabetes decisions. [10] Wearable HRV as a marker of inflammation is exploratory only. [11]

Who should be careful

Anyone who finds that watch data raises worry or compulsive checking should consider a break; see Tracking Anxiety and Compulsive Measurement. Do not change medicines because of a watch reading; talk with your prescriber. Chest-strap and clinic devices are compared in Chest Strap vs Wrist Heart-Rate Tracking, and the broader overview is Wearables and Health Trackers.