Your Pokémon Go Data is Being Used to Train AI Robots

When Pokémon Go launched in 2016, it wasn’t just a mobile game; it was a global phenomenon. From the parks of London to the plazas of New York, it seemed every smartphone owner was wandering the streets in a digital scavenger hunt. Although Niantic eventually offloaded the title to Scopely in March 2025, the colossal archive of geospatial data gathered during those peak years has found a surprising new purpose: teaching robots how to see.

This development follows an announcement from Niantic Spatial regarding a new partnership with Coco Robotics, the creators of an autonomous urban delivery fleet. Through this collaboration, Coco Robotics will utilize Niantic’s “spatial AI” and advanced Visual Positioning System (VPS) to sharpen the navigational precision of its “Coco” droids. These boxy, sidewalk-roaming robots—currently deployed in major hubs like Los Angeles, Chicago, and Helsinki—are designed to transport groceries and pizzas. However, they often struggle with signal degradation in dense city centers, where towering architecture interferes with traditional GPS.

“The potential for last-mile automation is vast, but successfully traversing the unpredictable chaos of city sidewalks is a monumental engineering feat,” stated John Hanke, CEO of Niantic Spatial, in a company update. Hanke, who established this AI-centric branch in May 2025, noted the striking parallels between gaming and robotics. He remarked that the logic required to make a Pikachu interact naturally with the environment is essentially the same logic needed to guide a Coco robot safely through a crowded thoroughfare.

Image: Niantic/The Pokémon Company

The roadmap for these robots is built upon years of data submitted by players of Pokémon Go and its precursor, Ingress. By leveraging VPS, Niantic can pinpoint a device’s location based on visual surroundings rather than a satellite ping. By encouraging millions of users to record 3D scans of landmarks from various angles and under diverse weather conditions, Niantic effectively crowdsourced a high-fidelity map of the world’s most complex urban environments.

In 2020, Niantic introduced “Field Research” tasks that rewarded players with rare items and Pokémon in exchange for these environmental scans. While the company has long been transparent about its intent to build a planetary-scale AR map, some trainers may be surprised to learn that their quest for a high-stat monster was actually a data-gathering mission for fast-food logistics. Whether fans will embrace their role as unwitting robotics trainers is unclear, but many are simply hoping the Coco fleet has a better sense of direction than the infamously lost Champion Leon.

Source: Polygon