Every irregular-rhythm feature on your wrist today answers the same question. Is something happening right now, or did something happen earlier that we noticed late?
A newly published Google patent application goes after a different question entirely. When is it going to happen again?
The application is numbered US 2026/0294263 A1 and was published at the start of October 2026, according to early coverage of it. At the time of writing the full text has not yet indexed into the public patent viewers, so the mechanism described below comes from two independent write-ups of the filing rather than the filing's own page.
What the Patent Actually Describes
The setup is straightforward for a patent in this space. A wearable detects an initial cardiac arrhythmia event using its own biometric sensors. Atrial fibrillation is the example the filing keeps returning to, though the claims are written broadly around cardiac arrhythmias.
That first event triggers an observation period. During the window, the watch keeps collecting heart data and records the additional events that follow. The system then looks at their timing and frequency to work out whether there is a pattern specific to that person.
If there is one, Google proposes using it to generate a prediction of when a future event may occur. Episodes that cluster at certain times of day, or that land at fairly regular intervals, are exactly the kind of signal that would feed that forecast.
That is a narrower idea than a general risk score. A risk score tells someone they are more likely to develop AFib at some point. This describes learning from the wearer's own recorded episodes and trying to call the next one.
Why This Is a Step Past What Ships Today
Current wearables can already flag AFib, and Google's devices are part of that group. The Fitbit Air uses heart rhythm and movement data collected while you are still to trigger an irregular rhythm notification. The Pixel Watch goes further with ECG hardware on supported models.
Both of those are still detection. They ask whether your rhythm looks irregular right now. Prediction flips it around and asks what might come next.
It also lines up with where Google has been pushing its health platform. Health Guardian already infers blood pressure and insulin resistance trends from signals the watch already collects, rather than from dedicated sensors that measure either one directly. AFib prediction follows the same logic. The sensors gather raw data, and more of the value comes from reading patterns across time.
We have also seen how that strategy lands in practice. The first Health Guardian reports came back confusing enough that users were not sure what to do with them, which is worth remembering when the next pattern-reading feature arrives.
The Part That Will Decide Whether You Ever See It
Patents are not roadmaps. Companies file on technologies that never reach a product, or that show up years later in a form nobody recognizes.
The bar is higher here than for a normal feature. Telling someone an arrhythmia may be approaching is a different promise from telling them one is happening. Detection is already a regulated health feature on these watches. A forward-looking warning would need convincing clinical validation before Google could push it to millions of wrists, and that is the sort of evidence a patent application does not contain.
There is a human problem too. Continuous rhythm monitoring can make people anxious, and a notification that says an episode may be coming has more room to cause alarm than one that reports what already occurred. Anyone who has watched a false positive on a heart feature knows how that goes.
Then there is the software question. A prediction needs a pattern, and a pattern needs repeated episodes. Someone who has had one recorded event has nothing to work with. That makes the feature useful for a specific slice of users, not for everyone wearing the watch.
What We Would Watch For
If this ever ships, the interesting details will be where it lands and how it is framed. Google runs both Fitbit and Pixel Watch lines, and health features have a history of arriving on one before the other. We have tracked how uneven that rollout gets, down to country-level availability.
The second detail is the wording. A prediction presented as a warning reads very differently from one presented as a heads-up. Detection features already lean on careful language for good reason.
For anyone who wants the current state of play on wrist-based rhythm detection, a Pixel Watch with ECG covers detection today, and Garmin's approach to irregular heartbeat notifications is worth comparing, which we have covered separately.
Bottom Line
Google has filed on the logic behind a prediction feature rather than a sensor. Its watches already have the hardware for detection, so the missing piece for forecasting is interpretation, which is exactly what the application describes.
That does not mean it is coming next year. It means the company is thinking past the notification that says something happened, and toward one that says something might. If that ever reaches a watch, the clinical evidence behind it will matter far more than the patent.
Sources: Gadgets & Wearables' write-up dated October 3, 2026, for the application number US 2026/0294263 A1, the initial-event-then-observation-window flow, the timing-and-frequency pattern learning, the AFib example and the "patents are not roadmaps" caveat; an independent German-language report from borncity, published early October 2026, confirming the same application number and the same mechanism of learning from the temporal patterns of actual episodes; and Google's own Fitbit and Pixel Watch heart rhythm support documentation for what irregular rhythm notifications and ECG currently cover. Two write-ups describe the application and both give the same number, but the application's own text was not yet available in the public patent viewers when this was written, so the number and the mechanism are attributed to that early coverage rather than to the document itself. No second independent write-up beyond these was found.
