Alphabet details how AI, 3D maps, and server-side processing aim to make Pixel Watch 5 GPS more accurate

Google has released a technical account of how it built what it calls the most accurate GPS route tracking yet for Pixel Watch 5, positioning the feature as a core part of the company’s broader smartwatch fitness pitch. The update is designed to help users retrace outdoor routes with fewer jumps or detours, including during runs, hikes, and bike rides.

At the foundation of the approach is how standard GPS works, and why it struggles in cities. GPS determines position by measuring how long it takes radio indicates from satellites to reach a device, then calculating location using indicates from multiple satellites. Google says that calculation can be thrown off by atmospheric conditions, nearby buildings, and satellite errors, leading to route lines that look unstable or wrong during workouts.

To address those urban distortions, Google’s Precision Location team built on earlier fixes used across Alphabet products, then modified them for Pixel Watch 5. A key element, according to the company, is combining Google Maps’ three-dimensional building models with AI, so the watch can better recognize how indicates may bounce off structures rather than arriving cleanly. Google also says it uses its network of GPS reference stations around the world, which measure GPS indicates, account for discrepancies, and provide that information to Google products in real time to reduce errors from the atmosphere and satellite anomalies.

Google’s blog post emphasizes that the company targeted real-world performance, not just lab results. Engineers reported verifying the approach through 4,000 miles of testing, and compared Pixel Watch 5 against other watches in what it describes as an internal study. In that evaluation, GPS performance was assessed based on route map accuracy in challenging urban environments for outdoor running, walking, and biking workouts under good network conditions.

The company names two engineers to frame the problem and the goal. Sharath Ananth, a lead engineer on Pixel Watch, describes accurate GPS routes as one of the pillars of a smartwatch’s fitness value. Frank van Diggelan, a distinguished engineer on the Precision Location team, says that if a watch’s position is “jumpy,” it undermines the entire outdoor activity experience.

Google says the hardest part was fitting the improved GPS system into the constraints of a smartwatch. Smartwatches use smaller antennas, which can limit the strength and quality of satellite reception. The company also points to power limits, noting that accurate route tracking requires handling a lot of data, especially when trying to compensate for weak indicates.

Google’s solution, as described in the post, is to compress raw GPS data on the watch and send it to a server for processing and correction. The corrected results are then delivered to the Google Health app backend. Google also states that Pixel Watch 5’s battery life was not negatively affected by this change, and that its performance was tested across different environments including open air and tree cover as well as dense urban areas.

To ensure the system worked globally, testers reportedly wore up to four different types of smartwatches and used a special backpack equipped with a massive antenna. The company says testers walked, ran, and biked thousands of miles across multiple cities and countries. In one internal review of results, an engineer questioned whether the data might be “leaking ground truth,” describing ground truth as control information representing the actual route a tester took. Google says that was not the case and that Pixel Watch 5 routes tracked closely enough to satisfy the team’s verification checks.

Google is now rolling the feature out commercially alongside Pixel Watch 5 itself. The company says Pixel Watch 5 will be available for preorder starting August 12, with sales beginning August 20. Pricing starts at $399 for the 41mm model and $429 for the 45mm model, and the watch is described as requiring an Android phone running Android 12.0 or newer, plus a Google Account and the Pixel Watch app.

What remains less clear from Google’s description is how much of the GPS improvement depends on conditions outside Google’s control, such as network quality, satellite geometry, and user behavior during workouts. The company does not disclose the specific AI model details, the size of the compression overhead, or exact error-rate figures, and it frames comparisons as internal tests using pre-production devices for sessions lasting three hours or less. Still, the combination of AI-assisted building-model correction and server-side processing suggests Google is prioritizing smoother, more reliable outdoor tracks where GPS reflection issues are most visible. Next, watch for broader third-party testing of Pixel Watch 5 GPS outdoors and for any follow-on changes to how Google Health or mapping features present those corrected routes to users.