The ECG button on the side of a smartwatch is one of the most over-explained and under-explained features in consumer electronics. The marketing copy makes it sound like you are wearing a cardiology clinic. The skeptical Twitter thread makes it sound like a placebo. Neither is right. A wrist ECG is a real medical signal, recorded by real electrodes, validated against real reference standards — and it is also a very narrow slice of what a cardiologist actually looks at. Knowing what slice you are getting is the entire point of this article.

What an ECG is, the short version

An electrocardiogram is a recording of the heart’s electrical activity over time. Every heartbeat starts as a small electrical impulse in the sinoatrial node, travels through the atria, pauses briefly at the atrioventricular node, and then sweeps down through the ventricles to produce the squeeze that pumps blood. Electrodes on the skin pick up the voltage changes from that traveling wave. Plotted against time, the waveform has a characteristic shape that cardiologists have been reading since Willem Einthoven first formalized it in the 1900s.

The familiar bumps have names. The small P wave is atrial depolarization, the upper chambers contracting. The big spike, the QRS complex, is ventricular depolarization. The T wave is the ventricles resetting electrically. The timing relationships between those features tell a clinician whether the rhythm is normal sinus, whether the atria and ventricles are talking to each other on schedule, and whether the conduction system is healthy.

A clinical 12-lead ECG records that signal from ten electrodes (six on the chest, four on the limbs) and produces twelve different electrical views of the heart. Different views show different problems. An inferior wall heart attack lights up leads II, III, and aVF. An anterior wall MI shows in V1 through V4. The reason cardiologists order a 12-lead is that one view is not enough to localize most things.

A smartwatch ECG records one view. Just one.

How the wrist hardware does it

The mechanics are simple and the result is genuinely a real ECG, not a derived estimate. The smartwatch has two electrodes that touch your skin. One is the metal back of the watch case, pressed against the wrist. The other is a metal contact you touch with a finger from the opposite hand: the digital crown on an Apple Watch, the side button on a Fitbit, the bezel on a Samsung Galaxy Watch, the crown on a Withings ScanWatch.

When you touch the contact, you close an electrical loop that runs from one wrist, up that arm, across the chest (and the heart), down the other arm, and into the watch. The two-electrode setup gives the watch the equivalent of clinical Lead I, the view between the left arm and the right arm. (A few devices, depending on how you hold them, produce what is closer to Lead II, the view from the right arm to the left leg. The Withings ScanWatch and some Garmin models can record either depending on contact placement.) The watch samples that voltage difference at a few hundred Hz for thirty seconds and produces the familiar waveform. That is the entire trick. The hardware is not exotic. The software is what does the heavy lifting after.

It is the same physical principle as a clinical ECG. It is just one lead instead of twelve, and the electrodes are imperfect compared to gel-backed clinical ones. Wrist contact is dry, the finger pressure varies, motion artifact is constant. The signal is noisier than a hospital tracing. For atrial fibrillation, which has a very characteristic look, the noise floor is low enough that it works.

What a single-lead ECG can actually detect

The single specific use case the FDA has cleared smartwatch ECGs for is screening for atrial fibrillation. AFib has a signature on any ECG lead, including Lead I: the rhythm is “irregularly irregular,” meaning the spaces between QRS complexes vary randomly rather than in a pattern, and the P waves (atrial activity) are absent or replaced by chaotic baseline wander. The R-to-R interval, beat to beat, has a distinctive variability that a classifier can learn to spot.

The reason AFib is the focus is partly that it is common (about 1 in 25 people over 60, much higher above 80) and partly that it is dangerous to miss. Untreated AFib roughly quintuples stroke risk. A lot of AFib is paroxysmal: it comes and goes, which means a single 10-second ECG in a clinic visit will often miss it. A wrist device that you have on most of the day, with passive irregular-rhythm detection from the optical heart-rate sensor and on-demand ECG you can take when you feel something off, fills a screening gap that clinics cannot fill. That is the clinical thesis behind the feature.

What the algorithms output is usually one of four labels: sinus rhythm (normal), atrial fibrillation (possible AFib), inconclusive, or low/high heart rate. The 30-second strip itself is saved as a PDF and is the part that actually matters. Your clinician can read it directly, and the algorithm label is secondary to a trained human looking at the waveform.

What it cannot do

This part is where most of the marketing misleads, and it is the reason for the callout at the top of this article. A single-lead ECG cannot detect most things a 12-lead ECG can. Specifically:

  • Heart attack and ischemia. MI shows up in ST-segment elevation or depression and pathologic Q waves, and the diagnostic pattern depends on which leads show the changes. A single lead cannot localize. None of the consumer smartwatch ECGs are FDA-cleared for MI detection. If you feel chest pain, the watch is the wrong tool.
  • Most arrhythmias other than AFib. Ventricular tachycardia, supraventricular tachycardia, atrial flutter, AV block of various degrees, long QT, Brugada syndrome. These have characteristic patterns that often require multiple leads to identify, and the algorithm will usually return “inconclusive” or misclassify them.
  • Structural heart problems. Chamber enlargement, hypertrophy, prior infarction scarring. These are read off the 12-lead morphology. Not visible on a wrist strip.
  • Anything happening when you are not taking a reading. The ECG itself is a 30-second snapshot. The passive irregular-rhythm notification (using the optical PPG sensor, not the ECG hardware) runs continuously, but it is less specific than the ECG and produces more false positives.

The irregular-rhythm notification deserves its own short note because people confuse it with the ECG. The notification is generated by the green-LED optical heart-rate sensor that is already running on your wrist all day. It looks at pulse-to-pulse intervals over multi-minute windows; if it sees the irregular-irregular pattern characteristic of AFib several times in a row, it raises a flag. That flag prompts you to take an ECG, which is the actual diagnostic-grade (well, screening-grade) recording. The notification alone is not an ECG, and it does not save a waveform. For more on how the optical sensor handles heart-rate variability in general, see our HRV explainer.

The Apple Heart Study, in actual numbers

The single most-cited piece of evidence for smartwatch AFib screening is the Apple Heart Study, published in November 2019 in the New England Journal of Medicine. The lead author was Marco Perez at Stanford, and the trial was Stanford-run but Apple-funded. Anyone making claims about smartwatch ECGs should know what it actually found, because the headline number gets misquoted in both directions.

The design: 419,297 Apple Watch users (Series 1 through 3, not yet with ECG hardware) enrolled remotely through the Apple Heart Study app. The watch’s passive irregular-rhythm algorithm watched their pulse. If a participant got a notification, they were mailed an ECG patch (a real one, the ePatch from BioTelemetry) to wear for up to seven days. The patch is the reference standard.

The findings, briefly:

  • 0.52% of the cohort received an irregular-pulse notification. About 2,000 of 419,000.
  • Of those who actually wore and returned the patch, the positive predictive value of a notification for any AFib on the patch was 0.84. Meaning if you got a notification and then wore a patch, there was an 84% chance the patch would catch at least one AFib episode during the recording.
  • But: only 21% of notified participants returned a usable patch. The denominator problem is real.
  • The cohort skewed young. Most participants were under 40. AFib prevalence in that age group is very low to start with, which inflates the apparent specificity.
  • The study was not designed to show clinical benefit (fewer strokes, longer life). It was a feasibility study showing that mass remote screening with a wrist device is possible.

The clean reading is this: a Series 3 Apple Watch, in 2019, on a self-selected population, with a passive PPG-based algorithm, could raise a hand and say “you should get this checked” with a real signal behind it about 84% of the time when the user actually followed up. That is a meaningful screening claim. It is not the same as a screening trial showing that wearing the watch saves lives. The follow-up Heartline study, with the VA and Johnson & Johnson, was designed to answer that question for older adults with the newer ECG-capable watches; it ran from 2020 through 2024 and has been releasing results in stages.

Which watches have FDA clearance, as of mid-2026

A non-exhaustive list of the consumer wrist devices with active US FDA 510(k) clearance for single-lead ECG, with the year of first clearance:

  • Apple Watch Series 4 and later (cleared 2018, K181837). Includes Series 4 through 10 and the Ultra line.
  • Fitbit Sense (2020) and Sense 2 (2022), plus Fitbit Charge 6 (2024). The Charge 6 is the first Fitbit tracker, rather than smartwatch, to ship with cleared ECG.
  • Samsung Galaxy Watch Active 2 and Galaxy Watch 3 and later (cleared 2020 in the US, K200073). The ECG feature was geographically gated for a long time and rolled out by region.
  • Withings ScanWatch (2019, after a delay) and ScanWatch 2 (2023). The original ScanWatch’s US clearance was famously delayed; the device shipped in Europe well before American buyers could record an ECG. See our ScanWatch 2 review for the long version.
  • Garmin Venu 2 Plus (2022) and select Fenix and Epix models. Garmin’s ECG rollout has been notably region-by-region. At one point the feature was available in the US but not the UK, and vice versa for some Fenix variants. Check Garmin’s regional product page before assuming the watch in your hand has it.

Pacemakers and ICDs are a contraindication on most of these; the manufacturer labelling tells you not to use the ECG feature with an implanted cardiac device. For a side-by-side of the actual sensor hardware across these watches, the wearable sensor accuracy explainer has the comparison table.

Who should actually buy one for the ECG

Our honest take, based on the clinical evidence and on reader emails over the past few years:

A smartwatch ECG is most useful for people who already have paroxysmal AFib diagnosed and want to keep a personal log of episodes between cardiology visits. Bringing a stack of PDF strips to a follow-up appointment is, in our experience, the way the feature actually pays off. The cardiologist can see the rhythm during a symptomatic episode, which an in-clinic ECG will rarely catch. Several cardiologists we have spoken to have started actively recommending wrist ECGs for exactly this reason.

It is also worth it for people with documented risk factors (prior cryptogenic stroke, age over 65, hypertension, sleep apnea, diabetes, heart failure, family history of AFib), particularly if you have noticed occasional palpitations. The screening signal is meaningfully better than not having a screening signal at all, and the irregular-rhythm notification covers you passively for most of the day.

It is less obviously worth it for healthy younger adults with no symptoms and no family history. The US Preventive Services Task Force concluded in 2022 (and reaffirmed it since) that there is insufficient evidence to recommend AFib screening with ECG in asymptomatic adults under 50, and the evidence is mixed even above that age. A wrist device will produce occasional false-positive notifications. Those notifications lead to clinic visits, patches, sometimes echocardiograms, and a non-trivial amount of anxiety. We are not telling you the watch is useless if you are 32 and healthy; we are telling you the ECG should not be the line item that decides which watch you buy. The HRV, sleep, and resting heart rate trends will give you more usable health information day-to-day than the ECG ever will. The best smartwatch for health tracking roundup is organized around that observation.

The other practical caveat is that the ECG is only as useful as your willingness to take it when you feel something. We have noticed Apple Watch wearers in our circle take ECGs more often than Withings owners, despite both having cleared hardware, simply because the Apple is already on the wrist when the palpitation happens.

For buyers choosing around ECG or heart-health features, start here:

A short note on the data itself

Every cleared device lets you export the ECG as a PDF. Apple does it through the Health app, Fitbit through the Fitbit app, Withings through Health Mate, Samsung through Samsung Health, Garmin through Connect. The PDFs are readable by any clinician who can read a Lead I tracing. Save the PDF. Email it to your doctor before the appointment. Do not rely on the in-app algorithm label as the only thing you bring; the waveform is the actual artifact.

We have read enough wrist ECGs at this point to have a working opinion: signal quality on the Apple Watch and Withings ScanWatch 2 is consistently the cleanest, the Fitbit Charge 6 is close behind, and the Samsung Galaxy Watch is fine but tends to have more baseline drift than the others in our hands. None of that is a substitute for what your cardiologist will tell you when they look at the strip themselves.