Google is working on predicting AFib before it happens

Google is looking at a fairly big next step for smartwatch heart tracking. Instead of simply telling you that an irregular heartbeat has happened, it wants a wearable to work out when another episode might be coming.

A newly published Google patent describes a system that monitors what happens after an initial cardiac arrhythmia event, looks for patterns in subsequent episodes and uses those patterns to predict when another event is likely to occur. The US patent application, numbered US 2026/0294263 A1, was published at the start of this month.

The watch would learn your own pattern

The basic idea is actually pretty simple. A wearable detects an initial cardiac arrhythmia event using its biometric sensors. The patent repeatedly uses atrial fibrillation as an example, although its claims cover cardiac arrhythmias more broadly.

That first event then triggers an observation period. During this window, the wearable continues collecting heart data and records additional arrhythmia events.

The system looks at their timing and frequency to see whether there is a pattern specific to that person. If there is, Google proposes using it to generate a prediction of when a future event may occur.

So this is not simply a general risk score telling someone they are more likely to develop AFib at some point. The patent describes learning from actual arrhythmia events and looking for temporal patterns that could help forecast another one. For example, if someone’s episodes repeatedly occur at certain times or tend to appear at fairly regular intervals, that information could potentially become part of the prediction.

This goes beyond today’s AFib alerts

Current wearables can already look for signs of AFib. Google’s own devices are part of that. I recently looked at how AFib detection works on Fitbit Air, where heart rhythm and movement data collected while the user is still can trigger an irregular rhythm notification.

Pixel Watch goes further by offering ECG hardware on compatible models. But these features are still fundamentally looking for something that is happening or has already happened.

Prediction flips that around. Rather than asking whether the wearer’s rhythm appears irregular now, the system starts asking what might happen next.

It also fits well with where Google has been taking its health platform lately. Health Guardian already uses existing wearable data to identify longer-term patterns linked to blood pressure and insulin resistance, rather than relying on dedicated sensors that directly measure either one.

AFib prediction would push that idea further. The sensors gather the raw signals, but more of the value comes from interpreting patterns across time.

But patents are not product roadmaps. Companies routinely patent technologies that never appear in a commercial device, or that arrive years later in a substantially different form.

There would also be a much higher bar for something that claims to predict a future cardiac event. Detecting signs of AFib is already a regulated health feature. Telling someone that another episode may be approaching would need convincing clinical validation before Google could put that sort of warning on millions of wrists.

Still, the direction is worth watching.