Your Smartwatch Caught an AFib Alert. Now What?

By A. Akotkar · 7 September 2026

Do smartwatch AFib alerts work? Yes. The sensors are genuinely good, and the pooled numbers are the kind manufacturers like quoting.

Then the awkward one. In the largest study ever run on these alerts, only about a third of the people who received one had atrial fibrillation on a follow-up ECG. Both facts are true at the same time, and the gap between them is not a flaw in the watch. It is arithmetic that almost nobody does.

This is not medical advice, and nothing here interprets anyone’s reading. It is a look at what the alert has been shown to mean across published studies. If your watch alerts you, the answer is a clinician, not a blog post.

TL;DR: Across 26 studies, smartwatches detect atrial fibrillation with 95% sensitivity and 97% specificity. In the Apple Heart Study, of 450 people who got an alert and returned a usable ECG patch, 34% had AFib. A 97%-specific test still produces mostly false alerts in a population where the condition is rare, and most watch wearers are that population. The US Preventive Services Task Force still says the evidence is insufficient to recommend screening asymptomatic adults either way.

• Do smartwatch AFib alerts work?
• Why are most alerts still wrong?
• What is the difference between an alert and a diagnosis?
• Should you be screening yourself at all?
• What does a false alert cost?
• So what should you actually do?

Do smartwatch AFib alerts work?

At the thing they measure, yes, and the evidence is not thin.

A systematic review and diagnostic meta-analysis published in JACC: Advances in November 2025 pooled 26 studies covering 17,349 patients. Overall sensitivity was 95% (95% CI: 92–97) and specificity 97% (95% CI: 94–98). The protocol was pre-registered with PROSPERO, which is worth noting because a lot of wearable research is not.

Read what that measures, though. Sensitivity and specificity describe how the device performs against a reference ECG, in a study population, with the answer already known. Neither number tells you what a specific alert on your own wrist means. That is a different question with a different answer, and it is the one everybody actually has.

Why are most alerts still wrong?

Because the condition is rare in the people wearing the watches, and rarity beats accuracy.

The Apple Heart Study, published in the New England Journal of Medicine in 2019, remains the largest test of this in ordinary use. It recruited 419,297 participants and monitored them for a median of 117 days. Just 2,161 — 0.52% — ever received an irregular-pulse notification. Of the 450 who returned an ECG patch with analysable data, atrial fibrillation was present in 34% (97.5% CI: 29–39). Among those aged 65 and over it was 35%.

So roughly two in three people who got an alert, and went to the trouble of wearing a patch, had no AFib recorded on it.

The arithmetic behind that is worth seeing, because it is not intuitive. Take the pooled 95% sensitivity and 97% specificity above and apply them to 100,000 people:

• At 0.5% prevalence: about 475 true alerts against 2,985 false ones — roughly 1 in 7 alerts is real.
• At 1%: about 950 true against 2,970 false — roughly 1 in 4.
• At 5%: about 4,750 true against 2,850 false — roughly 5 in 8.

That calculation is mine, from the published sensitivity and specificity, and it is illustrative rather than a finding: real algorithms require repeated detections rather than firing on a single reading, which pushes performance better than the raw numbers suggest. The direction is the point. A 97%-specific test aimed at a rare condition produces mostly false positives, and it is not broken when it does.

Two honest caveats on the Apple figure. The patches went on an average of 13 days after the alert, and atrial fibrillation comes and goes — an absent patch reading is not proof the alert was wrong. And the study was funded by Apple, which is disclosed in the paper and worth carrying forward rather than dropping.

What is the difference between an alert and a diagnosis?

One is a pulse-timing sensor noticing irregularity. The other is a recording of the heart’s electrical activity, read by someone qualified to read it.

Optical sensors work by photoplethysmography: a light shone into the wrist, measuring the timing between pulses. Irregular timing is a reason to look closer. It is not, on its own, atrial fibrillation. A 2026 letter in JACC: Advances responding to that same meta-analysis makes the point in its title — “Smartwatch Accuracy for Atrial Fibrillation: Do Not Confuse Alert and Confirmation”.

This is also where the most-quoted number gets misread. The Apple study reported a positive predictive value of 0.84 (95% CI: 0.76–0.92), and that figure travels around as “84% of alerts are correct.” It is not what the paper measured. It is the probability of AFib appearing on the ECG simultaneously with a subsequent irregular-pulse notification — a narrower claim about the device agreeing with itself while a patch happened to be recording. The number for “had AFib on the patch at all” is the 34%.

Should you be screening yourself at all?

Nobody knows, and that is the official position rather than a hedge.

The US Preventive Services Task Force gives AFib screening in asymptomatic adults aged 50 and over an “I” statement: the evidence is insufficient to assess the balance of benefits and harms. Its 2022 update explicitly widened scope to consumer wearables and lowered the age from 65 to 50, and still landed on insufficient.

An “I” statement is not a recommendation against. It means the trials that would settle whether finding asymptomatic AFib this way prevents strokes have not been done. Detecting more of something is only useful if acting on it changes outcomes, and that link is the part still missing. It does not apply to people with symptoms, or with a history of stroke or TIA — different situation, different evidence.

What does a false alert cost?

More than nothing, which is the assumption worth challenging.

In the Apple study, of 1,376 notified participants who returned a 90-day survey, 57% contacted a health care provider outside the study. That is over half of a group in which roughly two thirds turned out not to have AFib on a patch.

There is some evidence the alerts themselves do harm. The Pulsewatch study, which randomised patients aged 50 and over with a history of stroke or TIA to a smartwatch plus patch or a patch alone for 14 days, found that receiving false AFib alerts was associated with a dose-dependent decline in self-perceived physical health and in confidence managing chronic symptoms. Treat that as suggestive rather than settled: it is a small study in a low-profile journal, and it measured self-reported outcomes over two weeks.

The counterweight is real too. Atrial fibrillation genuinely raises stroke risk, it genuinely can go unnoticed, and a watch that prompts someone to get checked has obvious value in the cases where it is right. The point is not that the alerts are useless. It is that “more detection” and “better outcomes” are separate claims, and only the first one has been demonstrated.

So what should you actually do?

I have not tested any of these devices — no watch, no patch, no comparison. This is a reading of the published studies, not a hands-on review, and it is not a substitute for a clinician. What the evidence supports:

• Treat an alert as a prompt to get checked, not as a finding. The device has noticed irregular pulse timing. Confirming atrial fibrillation takes an ECG and someone qualified to read it.
• Do not treat silence as reassurance. Sensitivity is not 100%, the watch is not measuring continuously, and symptoms matter regardless of what the device says.
• Your own risk changes what an alert means. The same alert carries far more weight at 75 than at 30, because the arithmetic above turns on how common the condition is in people like you.
• Don’t buy one specifically to screen yourself. The USPSTF position is that nobody yet knows whether screening asymptomatic people this way helps.
• Expect false alerts, and expect them to be unpleasant. Roughly two thirds of alerts in the largest study did not correspond to AFib on a patch, and there is some evidence the experience takes a toll.

For the other end of the same question — what movement actually does for cardiovascular risk, where the evidence is much stronger — see whether exercise offsets a day of sitting.