Can a Smartwatch Detect AFib? What the Apple, Fitbit and Samsung Evidence Actually Shows

Short answer: yes, a smartwatch can detect atrial fibrillation, using two different tools with different strengths.

• A passive alert. The optical heart rate sensor checks pulse regularity in the background and notifies the wearer when it sees a pattern suggestive of AF. Apple, Fitbit and Samsung all have FDA clearance for this [1][2][3].
• An on-demand ECG. The wearer touches the crown or bezel for 30 seconds and the watch records a single-lead electrocardiogram. Under European guidelines, a physician reading 30 seconds or more of single-lead ECG showing AF can make the diagnosis from it [4].

AF is the most common sustained arrhythmia and a major cause of stroke, and it is often silent. That is why the heart studies behind these features enrolled hundreds of thousands of people, and why cardiologists, health plans and remote monitoring programs care what a notification means. This post covers what the big studies found, why two meta-analyses disagree, what the newest randomized trial showed, what guidelines accept, and what a product should do with the data.

What did the big smartwatch heart studies find?

Apple Heart Study (2019). 419,297 participants without known AF used an Apple Watch irregular pulse algorithm for a median of 117 days. 0.52% received a notification. Notified participants were mailed an ECG patch, which arrived on average 13 days later. Among the 450 who returned usable patches, 34% had AF on it, 35% of those 65 and over. When a notification fired while the patch was being worn, it was AF 84% of the time [5].

Fitbit Heart Study (2022). 455,699 participants using Fitbit devices, median age 47. The algorithm required 11 consecutive irregular five-minute windows before alerting. 1% were notified, 4% of those 65 and over. 32.2% of notified participants with a usable patch had AF. When an alert fired during patch wear, 98.2% were AF [6].

Huawei Heart Study (2019). 187,912 people in China, mean age 35. 0.23% received a suspected AF notification, and 87% of those followed up were confirmed by clinical evaluation, ECG or Holter monitoring. About 80% of high-risk patients with confirmed AF were started on anticoagulation through an integrated care pathway [7].

The gap between the two Apple and Fitbit numbers is the most important thing in the chart. When the watch alerts, it is usually right. A week later, AF is often gone. Much early AF is paroxysmal: it comes and goes. A patch that goes on two weeks after an alert can miss it, which is why “only a third had AF” and “84% to 98% were right” are both true.

Why do the meta-analyses disagree?

Two recent pooled analyses give quite different accuracy figures.

Three things explain most of the gap.

• Which tool is being tested. ECG classification in controlled settings tends to score higher than passive PPG monitoring in daily life. Barrera found PPG and ECG models comparable across its included studies; the broader Apple review pools more real-world designs [8][9].
• What happens to unreadable recordings. Many ECG validation studies report accuracy only on recordings the algorithm could classify. An inconclusive result is neither a hit nor a miss in those numbers, but it is a real outcome for a user.
• Who is enrolled. Studies in cardiology clinics, with many patients known to have AF, produce very different numbers from population screening in mostly healthy people.

The practical reading: under good conditions a smartwatch ECG classifies AF well. In everyday passive monitoring, sensitivity is lower and depends on how often and how long the watch is worn.

Does finding AF with a watch help anyone?

This is the question that decides whether payers and health systems should pay attention, and the evidence is still arriving.

EQUAL (2026). A randomized trial in the Netherlands enrolled 437 adults aged 65 and over with elevated stroke risk and no known AF. Half used an Apple Watch with PPG alerts and ECG, reviewed by trained cardiac nurses within 24 hours, for six months [10][11].

• The watch group had 20 confirmed AF diagnoses; all diagnoses in the standard care group were symptomatic.
• That was a 7.3 percentage point absolute increase, a number needed to screen of 14.
• 57% of the watch-detected AF was asymptomatic.
• Everyone diagnosed was started on anticoagulation.
• The positive predictive value of the first AF-labeled watch ECG was 54%, which is why nurse review mattered.

The trial was not designed to show fewer strokes. Its authors say the clinical impact of treating screen-detected AF still awaits outcome trials [11].

Heartline. A decentralized randomized trial by Johnson & Johnson and Apple in more than 34,000 US adults aged 65 and over, testing whether Apple Watch features plus an engagement program lead to earlier AF diagnosis and better outcomes, including stroke [12]. It was presented as a late-breaking trial at the American College of Cardiology meeting on 30 March 2026. We could not find a peer-reviewed publication of its outcome results as of this writing.

The reason caution persists is history. Earlier screening trials using patches, implanted loop recorders and intermittent ECGs, including mSToPS, STROKESTOP and LOOP, consistently found more AF, but more detection did not consistently reduce clinical outcomes [12].

What do guidelines say?

US screening. The US Preventive Services Task Force concluded in 2022 that current evidence is insufficient to assess the balance of benefits and harms of screening for AF in adults 50 and older without signs or symptoms [13]. That is not a recommendation against watches. It is a statement that the outcome evidence is not there yet.

European diagnosis. The 2024 European Society of Cardiology guideline states that a single-lead ECG of 30 seconds or more showing the characteristic rhythm is diagnostic of clinical AF, and recommends that a physician review the ECG, whether 12-lead, multi-lead or single-lead, to confirm the diagnosis [4]. That places smartwatch ECGs inside the diagnostic pathway. PPG alerts on their own are not.

AF burden as an endpoint. In 2024 the FDA qualified Apple’s AFib History feature as a Medical Device Development Tool, the first digital health technology qualified this way, so it can be used to estimate AF burden in clinical studies [14]. For people already diagnosed, the watch has become a measurement instrument, not only an alarm.

What should a care program or product do with AF data?

• Route alerts to review, not to the patient alone. EQUAL’s nurse review within 24 hours is the model: a 54% positive predictive value on the first ECG label means someone qualified needs to look [11].
• Ask for an ECG at the moment of an alert. The heart studies show the alert is most informative when it happens. A watch ECG recorded then is worth more than a patch mailed later.
• Focus on people with stroke risk. Detection matters because anticoagulation reduces stroke in people with AF and risk factors. That is where the EQUAL and Heartline populations sit, and where Medicare Advantage programs have the most to gain.
• Track burden for diagnosed patients. AF burden over weeks is more informative for management than a single alert.
• Stay inside cleared claims. Surfacing a device’s own FDA-cleared notification is different from inferring AF from raw heart rate yourself, which is a disease claim; see whether your health app is a medical device.
• Build for physicians’ workflow. Rhythm data is among the wearable data physicians review most, and still rarely reaches the chart; see how doctors use wearable data.

Where does AF data live?

On iPhone, apps can read it with permission. HealthKit exposes irregular heart rhythm events from iOS 12.2, ECG recordings including voltage data and Apple’s classification from iOS 14, and AFib burden from iOS 16 [15][16][17].

On Android, it mostly stays with the vendor. Health Connect has heart rate and HRV types, but no ECG or AF type, so Samsung’s and Google’s rhythm results stay in their own apps [18]. A product that wants rhythm data across platforms has an iPhone path and, on Android, depends on vendor partnerships.

Raw heart rate is not a substitute. Minute-level heart rate in either store is too coarse to detect AF; the cleared algorithms work on beat-to-beat intervals the platforms do not generally expose. The limits of wrist heart rate itself are covered in resting heart rate accuracy.

The short version

Smartwatches detect AF two ways: passive optical alerts and on-demand ECGs. In the Apple and Fitbit heart studies, alerts were AF 84% to 98% of the time when they fired, but only about a third of alerted people showed AF on a patch a week later [5][6]. Meta-analyses put accuracy anywhere from about 80% to 95% sensitivity depending on the tool and setting [8][9]. A randomized trial in high-risk older adults found about four times as many diagnoses with a watch over six months, mostly of silent AF [10][11], and guidelines now accept a 30-second single-lead ECG for diagnosis [4], while US screening recommendations wait for outcome data [13]. The useful product is fast clinical review of alerts in people at risk, not the alert itself.