Apple Watch heart rate is accurate at rest and steady effort
High confidence
On average the Apple Watch reads your heart rate within about a beat of a chest strap or ECG, at rest and on a run. Any single reading can be off by about 7 beats either way, and errors run larger when the wrist is busy, as in lifting, and tend to grow as effort climbs.
Why it works
The watch uses photoplethysmography (PPG), shining light into the wrist to read how blood volume shifts with each beat. Movement, moisture, and loose skin contact blur that signal, which is one reason readings slip during hard efforts with irregular wrist motion.
The evidence
-
This living review pooled 82 studies covering 430,052 participants and 14 health metrics measured by Apple Watch through Series 9 and Ultra 2. Heart rate showed a small mean bias (-0.27 bpm) with moderate variability (limits of agreement about -7 to +7 bpm), and during exercise 10 of 11 studies (91%) reported a mean absolute percentage error under 10%. Energy expenditure was the weakest metric: all six studies that calculated mean absolute percentage error reported 20% or higher in at least one condition, with values ranging from about 9.7% (running) to 151.7% (walking). Blood oxygen bias was near zero (-0.04%) but with wide agreement limits, and atrial-fibrillation detection had 79% sensitivity and 91% specificity, so a notification likely reflects true atrial fibrillation and warrants clinical follow-up, though many studies had more than a quarter of measurements come back inconclusive.
-
In 39 healthy adults who gave 316 measurements over 14 days, Apple Watch Series 9 and Ultra 2 heart-rate-variability (HRV) readings were compared against a Polar H10 chest strap (the reference). The watch underestimated HRV by an average of 8.31 ms (p = 0.025), with a mean absolute percentage error of 28.88% and a mean absolute error of 20.46 ms, and it did not fall within the pre-specified ±10 ms equivalence margin; the authors say further refinement is needed to reach clinical-level accuracy. Resting heart rate, by contrast, was accurate (mean difference -0.08 bpm, MAPE 5.91%). The authors urge caution in interpreting watch HRV for clinical or performance monitoring.
n=39
-
Wrist-worn optical heart rate showed small mean differences during sleep, rest, and treadmill walking-to-running, but larger underestimates during resistance training and cycling. In resistance training the error grew as heart rate rose. The authors judge wrist-worn devices acceptably valid for most common activities.
-
Apple Watch heart rate showed a near-zero mean bias with wide limits of agreement, and no heart rate subgroup exceeded the 10% MAPE validity threshold. Energy expenditure exceeded that threshold in every subgroup. Accuracy varied across user characteristics and measurement conditions.
Why we call confidence high
Two Apple Watch meta-analyses agree: a 2026 living systematic review (Lambe et al., 82 studies) found a pooled heart rate bias of -0.27 bpm with limits of agreement of about -7 to +7 bpm, and 10 of 11 exercise studies under 10% error; a 2025 meta-analysis (Choe and Kang, 56 studies) found a bias of -0.12 bpm and no heart rate subgroup above the 10% error threshold. Lambe reports error tended to rise with intensity (three studies saw the reverse), and a 44-study meta-analysis of wrist-worn optical sensors (Zhang 2020) found larger error during resistance training and cycling than at rest or on the treadmill. O'Grady 2024 found resting heart rate accurate on Series 9 and Ultra 2 but HRV error (~29%) too high for clinical use. Falter 2019, in cardiac patients cycling, found agreement improved at peak effort, so the intensity clause is a tendency, not a rule.
Where it applies
Adults, mostly healthy, in lab validation studies (treadmill, cycling, resistance exercise, rest)
Last reviewed Oct 14, 2026. See how we score.