Which Smartwatch Health Claims Actually Hold Up in 2026?

People make real decisions with wrist data now. They change how they train because a readiness score dropped, worry about a night the watch labelled low on deep sleep, or read an ovulation window as a fact rather than a guess. So the question worth asking is not which smartwatch is best. It is which of the numbers on the screen are accurate enough to act on.

We went through the published validation literature rather than the marketing pages, because these features vary enormously in how well they have been tested. Some have regulatory clearance and peer-reviewed trials behind them. Others are proprietary scores with no published validation at all. Knowing which is which changes how you should read your own data.

One thing to say plainly up front: this is a research summary, not medical advice, and we are not clinicians. Nothing here replaces a conversation with your doctor.

Is wrist heart rate accurate during exercise?

It depends entirely on the exercise. Optical wrist sensors work by shining light into the skin and reading how blood flow changes the reflection, which requires the sensor to stay still against the wrist. During rest and steady-state cardio like walking or cycling, published comparisons against chest straps and ECG generally find wrist sensors close to reference equipment, with small error margins.

Resistance training and interval work are the problem. Wrist flexion, grip tension and rapid heart rate changes all degrade the signal, and validation studies consistently report substantially wider error during these activities than during steady cardio. That has a practical consequence. If you are pacing a run or tracking resting heart rate over months, wrist data is fine. If you are reading a calorie figure from a lifting session, you are reading an estimate built on an unreliable input.

Anyone who wants dependable heart rate during hard or variable training should use a chest strap, which measures electrical activity rather than optical blood flow and does not care how much your wrist moves.

Can a smartwatch really track your sleep stages?

Total sleep time, reasonably well. Sleep stages, no. The clinical standard for staging is polysomnography, which uses EEG to read brain activity along with eye movement and muscle tone. A watch has none of that. It infers stages from movement and heart rate variability, and validation studies comparing consumer wearables against polysomnography repeatedly find that stage classification, particularly separating REM from deep sleep, falls well short of clinical accuracy.

Read the summary rather than the breakdown. Whether you slept roughly seven hours or roughly five is genuinely useful information, and the trend across a month is more useful still. A specific claim that you got 42 minutes of deep sleep is a modelled guess presented with false precision, and it is not worth losing sleep over, which is the irony several sleep researchers have pointed out about this category.

Which smartwatch health features have real regulatory clearance?

This is the sharpest line in the category, and it separates a handful of features from everything else. ECG for atrial fibrillation detection on devices like the Apple Watch and Withings ScanWatch has FDA clearance, and the Apple Heart Study was published in the New England Journal of Medicine. Some devices also carry clearance for sleep apnea notifications. Clearance means a regulator reviewed evidence for that specific claim.

Most features have nothing comparable. Stress scores, readiness and recovery scores, energy budgets and body battery metrics are proprietary composites, and brands generally do not publish the validation behind them. They may still be useful as personal trend lines, since a score that drops when you are genuinely run down is telling you something. It is simply not a measurement in the way an ECG reading is.

Wrist body composition works by bioelectrical impedance, sending a tiny current between contact points. Hydration, recent meals and skin contact all shift the reading, so treat the output as a rough trend rather than a body fat percentage you would quote to anyone.

Are smartwatch cycle tracking predictions reliable?

Cycle prediction has improved, and temperature sensing on newer devices gives the algorithms a genuine physiological signal rather than calendar math alone. Basal body temperature does shift after ovulation, so a watch tracking overnight skin temperature has something real to work with.

The limitation matters more than the improvement. These predictions are retrospective and probabilistic, they are not a substitute for ovulation testing or bloodwork, and no consumer wearable diagnoses a hormonal condition. If you are trying to conceive, or you suspect something like PCOS or thyroid dysfunction, the watch is not the tool for that answer. Reassuring app output has led people to postpone a proper workup, which is the genuine harm in this category.

How should you actually use the data?

Trends beat single readings. One night of poor sleep or one elevated resting heart rate reading means very little, while a resting heart rate that climbs steadily over three weeks is a real signal worth mentioning to a doctor. Consistency also beats accuracy for most purposes. A device that reads slightly low every night still shows you change over time, which is what you actually want from it.

A reasonable way to read your own dashboard: step counts and resting heart rate are solid, total sleep duration is decent, workout heart rate is good for cardio and rough for lifting, sleep stages are entertainment, and proprietary wellness scores are unvalidated composites. ECG is the one feature that carries genuine clinical weight, on the devices cleared for it.

So which smartwatch should you buy?

Match the device to the claim you care about. Anyone with cardiac risk factors or a family history of atrial fibrillation should buy a device with cleared ECG, because that is the feature with real evidence behind it and it is worth paying for. Endurance athletes usually favour Garmin for battery life and training metrics, though serious training still deserves a chest strap. Anyone buying a first wearable to establish baselines does not need a flagship, since step counts and resting heart rate come from the same sensor class on a budget tracker.

What no smartwatch does is diagnose. It flags patterns worth investigating, which is genuinely valuable, and the investigating still happens somewhere with actual lab equipment. Symptoms deserve a doctor whatever your wrist says, and a reassuring number is not a clean bill of health.