When Pokémon Go first launched in 2016, it felt like pure magic. The game sent players outside, into parks and city streets, transforming neighborhoods into digital battlefields and PokeStops. It wasn’t just a game—it was a movement. But as it turns out, it wasn’t just about catching Pokémon.
Fast forward to current times, and revelations have emerged that Pokémon Go players have been doing more than simply gaming. They’ve been feeding an artificial intelligence—one that doesn’t just understand the world but learns from it, much like a human. The game’s developer, Niantic, has been quietly constructing what it calls a Large Geospatial Model (LGM)—an AI-powered map that could one day power the next generation of augmented reality (AR), robotics, and even military systems.
To put it bluntly: while you thought you were training your Pikachu, you were actually training Niantic’s AI to understand the world.
How Niantic Turned Pokémon Go Into a Stealthy AI Training Ground
If you’re not familiar with Pokémon, the franchise began in the 1990s with a Japanese animated series (anime). Nowadays, Pokémon spans video games, trading cards, TV shows, and movies. The core idea is that players, known as Trainers, capture and train fictional creatures called Pokémon. The anime’s protagonists often fight against Team Rocket, the fictional antagonistic syndicate.
With Pokémon Go, Niantic brought the beloved franchise into the real world, allowing players to catch, train, and battle creatures using their smartphones. The game was so successful that Niantic became known for pioneering location-based augmented reality (AR) experiences and basically set the bar for all subsequent projects.
Niantic, originally a part of Google, had previously experimented with AR mapping through its earlier game, Ingress, but Pokémon Go took the concept mainstream, merging nostalgia with cutting-edge technology. The game’s success turned city streets, parks, and landmarks into virtual arenas, encouraging social interaction and exploration. However, beneath its playful exterior, Pokémon Go was also laying the groundwork for something much bigger—a vast, AI-driven map of the real world, built unknowingly by its players.
Essentially, when players perform certain activities that offer information about the real world, that can become training data for the geospatial model.
An Announcement That Stunned Players
Niantic’s blog post announcing its LGM (Large Geospatial Model) explains the company’s AI ambitions. Much like how Large Language Models (LLMs) such as OpenAI’s ChatGPT have learned to process and generate human-like text, a Large Geospatial Model aims to do the same for the physical world.
This AI doesn’t just create a 3D model—it understands space. It can recognize the layout of a city street, differentiate between buildings and trees, and even anticipate what’s around the corner based on millions of other locations it has seen. Niantic’s Visual Positioning System (VPS), trained on over 50 million neural networks and 150 trillion parameters, is already operational in over a million locations.
So here’s the kicker: Pokémon Go players weren’t just collecting Pokémon. They were, in essence, unpaid cartographers for one of the most sophisticated AI mapping projects ever attempted. Granted, the LGM doesn’t gather all data from every Pokémon Go player at all times. Instead, it selectively collects data in specific instances—primarily through voluntary scans of real-world locations. Mashable describes how Niantic encouraged players to scan PokéStops, capturing high-resolution, GPS-tagged images of real-world locations.
Each scan added to a growing neural map—a detailed, AI-powered understanding of the world. Unlike traditional digital maps, which rely on street-level data from satellites and cars, this database comes from a pedestrian perspective. It captures back alleys, hidden parks, and places inaccessible to vehicles.
Is Pikachu training the military?
On the surface, Niantic’s project sounds groundbreaking; and it is. It could lead to smarter AR applications, better navigation tools, and even more immersive gaming experiences. It makes the game itself much better. But there’s a darker side to this story.
The data Pokémon Go players have generated could be used for far more than just gaming. Open Source Intelligence analyst Elise Thomas points out that the same AI models used to enhance AR glasses could also aid military surveillance or law enforcement tracking. If an AI can “see” the world as well as a human, it can also be used for security, monitoring, and even predictive policing.
And then there’s the question of consent. Did players realize they were training an AI when they signed up to catch Pokémon? Niantic emphasizes that all scanning was optional, but many players had no idea of the extent to which their gameplay was shaping AI models.
“We use player-contributed scans of public real-world locations to help build our Large Geospatial Model. This scanning feature is completely optional – people have to visit a specific publicly-accessible location and click to scan. This allows Niantic to deliver new types of AR experiences for people to enjoy. Merely walking around playing our games does not train an AI model,” Niantic writes.
So… Was Pokémon Company Team Rocket All Along?
Niantic’s LGM isn’t an evil scheme, but it does raise serious questions about data privacy, AI ethics, and the role of everyday people in training AI. Team Rocket, the infamous Pokémon villains, were known for using trickery to exploit Pokémon trainers for their own gains. Did Niantic do the same? It depends on how you look at it.
On one hand, players willingly participated in the game. They had fun. They explored their cities. On the other, they were unknowingly contributing to a global AI experiment—one that could shape the future of artificial intelligence, for better or worse.
Like a true Pokémon battle, the outcome depends on how Niantic chooses to use this powerful technology. Will it be used for good, powering the next generation of immersive AR? Or will it be co-opted for surveillance and control?
It’s clear now that Pokémon Go was more than just a game—it was an early test of something much bigger. Niantic envisions a future where augmented reality, AI-powered mapping, and geospatial intelligence merge to create a new digital landscape, one where computers don’t just store information about locations—they understand them.
