Millions of players helped create the world map that robots will use

But almost a decade later, another dimension is beginning to be seen behind that success: part of what Pokémon GO players did also helped build one of the three dimensional maps of the world most detailed ever created.

The company behind the game, Niantic, used scans made by users for years to develop spatial models capable of understanding what the physical world is like. That information is now beginning to be used for something very different from catching virtual creatures: helping robots and autonomous systems move with precision in cities.

When it appeared in July 2016, Pokémon Go became one of the biggest cultural phenomena of the decade. The game used the location of the cell phone and the camera to superimpose creatures from the Pokémon universe over the real environment.

The proposal was simple: walk around the city to find Pokémon, visit “pokéstops” in emblematic places and participate in gyms where battles were fought.

During the first months, scenes of players walking in groups through parks or squares were repeated around the world. According to the company’s own data, more than 500 million people downloaded the game in its first years, making it one of the most popular mobile titles in history.

But another key technology was also hidden behind these mechanics: capturing images of the real environment.

Over time, the game began to include an optional feature that invited players to scan specific locations, such as PokéStops or gyms, with the camera.

The process was simple: the user had to walk around the object or monument while recording a video of a few seconds, generally with 180 or 360 degree movements.

As Niantic explains, these scans were automatically anonymized and the system blurred identifiable elements like faces or license plates. But the data was extremely valuable.

Each capture included information such as exact location, camera angle, direction of movement, time of day, and lighting conditions.

With millions of players performing these actions in different parts of the world, the result was unexpected: a huge visual database of the planet.

For years, Niantic explained that this data would be used to improve the augmented reality experience within the game itself. However, the goal was more ambitious.

In 2024, the company presented its Large Geospatial Model, an artificial intelligence system capable of understanding physical spaces from millions of geolocated images.

That model was trained with more than 30 billion images taken in millions of places around the world, many of them captured by players while using Pokémon Go.

The logic is similar to that used by language models in artificial intelligence: the more material is analyzed, the more capable the system becomes of recognizing patterns.

In this case, the model learns what cities look like from different angles, at different times of the day and under different weather conditions.

The story took another turn in 2025. Niantic sold its video game business to Scopely and spun off its technology division into a new company called Niantic Spatial.

That company presents itself today as a firm specialized in geospatial artificial intelligence, that is, technology capable of interpreting the physical world.

The next step came in March 2026, when Niantic Spatial announced an agreement with Coco Robotics, a company that operates delivery robots in cities in the United States and Europe.

The idea is use the visual positioning system developed from game data to improve the navigation of these robots.

The problem it tries to solve is known: GPS works well in open spaces, but in cities with tall buildings it can fail by tens of meters.

For a person, that margin of error is smaller. Just look around to find your way. But for a robot that must stop exactly in front of the door of a restaurant or a house, that inaccuracy can be critical.

That’s where the three-dimensional map generated from millions of scans comes into play.

The idea of ​​using users’ everyday actions to train technology is not new. One of the best known examples are the CAPTCHAthose tests that ask you to identify traffic lights, bicycles or pedestrian crossings before entering a website. These exercises helped for years to train the artificial vision systems used by autonomous cars.

The difference is that in the case of Pokémon Go The scale was gigantic..

While projects like Google Street View They depend on cars equipped with cameras that travel the streets, the game achieved something different: thousands of photographs of the same place taken from different angles, at different times and with multiple lighting conditions.

In other words, an urban scan distributed among millions of people.

For Niantic, the project demonstrates that video games can become an unexpected source of data for the development of new technologies. But it also raises questions.

Many players believed they were only helping to improve the game’s augmented reality. Today it is clear that this information also fed systems that they can end up in robotsurban logistics or future augmented reality devices.