Every year for the past three, somebody has predicted that AI was about to kill geography games. The argument is straightforward: if a vision model can identify a country from a single photograph, what is the point of a human guessing the same country from the same photograph? The skill is now trivial. The genre has been automated out of existence. Time to move on.
The actual data tells the opposite story. The audience for geography games has grown faster in the past two years — exactly the period when AI image recognition has been improving most rapidly — than it did in the five years before. The number of people playing daily geography puzzles, satellite imagery games, and street-view guessers is at an all-time high. The most popular geography YouTube channels have grown their audiences while AI has gotten better, not worse. Something about the "AI kills geography games" prediction is wrong. This article is about what.
Geography Games Were Never About Finding the Right Answer
The core mistake of the "AI will replace it" argument is treating geography games as if they were lookup tasks. They are not. If finding the right answer were the only point, you could already use a reverse image search to identify almost any GeoGuessr round with reasonable accuracy. People still play. The audience is not playing because they want to know where the round is. They are playing because they want to reason their way there.
This is the same reason chess engines did not kill chess. Stockfish has been able to beat the world champion for decades. Chess.com is bigger than it has ever been. The audience plays not because they need to know the best move — they could ask the engine — but because they want to find the best move themselves. The pleasure is in the doing, not the answer.
What AI Actually Changes
AI does change the geography game landscape, but not in the way most people expect. It mostly changes what counts as an interesting round. The trivial rounds — instantly recognisable landmarks, famous coastlines, obvious cities — become less interesting to play, because the answer is essentially looked up. The harder rounds — featureless deserts, anonymous taiga, ambiguous coastlines — become more interesting, because they are exactly the kinds of rounds where AI still struggles and where human reasoning still has a clear edge.
Game designers respond to this naturally. Modern geography games tend to emphasise harder location sets, larger maps with more obscure territory, and scoring systems that reward consistent skill across the difficulty range rather than perfect performance on the easy rounds. The genre is evolving toward where the human edge is greatest. AI has not eliminated the challenge — it has shifted the challenge to the parts of the world that are most worth getting better at.
AI as a Coach, Not a Competitor
The most interesting AI applications in geography games are not about replacing the human player. They are about coaching. After you guess, an AI can tell you what specifically was visible in the round that should have led you to a better answer — "the herringbone deforestation pattern is characteristic of Brazilian Amazon roadside settlement," or "the specific colour of the soil here is associated with the loess plateau in north-central China." This kind of post-round commentary is genuinely valuable, and a few products are starting to explore it.
Played well, this turns AI into an accelerator of human learning rather than a substitute for it. The same way a chess engine has become an indispensable training tool for chess players (they review their own games against the engine's analysis), AI vision models are becoming an indispensable training tool for serious geography players. The player's actual play is still human; the analysis afterwards is AI-augmented; the resulting skill is human, deepened by AI commentary.
Why the Audience Has Grown
The deeper reason geography games have grown in the AI era is that AI's broader effect on the world has made human skills feel scarcer and more valuable. When everything routine is being automated, the activities that train deliberately human cognition — pattern recognition under uncertainty, spatial reasoning, slow careful observation — become more interesting, not less. Geography games sit squarely in that bucket. They are explicitly about looking carefully at the physical world and reasoning your way to a conclusion. That kind of practice is exactly what people increasingly feel they are losing in their automated daily lives, and it turns out a daily two-minute habit can quietly rebuild it.
There is also the simple reality that the world is more interesting when you can see it more clearly. The geographic literacy that geography games build is its own kind of compounding pleasure — once you can see the world more sharply, you keep seeing it more sharply. AI does not threaten this. If anything, it sharpens it.
The Long View
It is possible to imagine a future where AI has so thoroughly solved the recognition task that even the hardest geography rounds are trivially identified. That future is not close — current models still struggle badly on featureless rounds — but suppose it arrives. The question becomes: do people still want to play? The honest answer, based on every parallel we have (chess, music performance, mathematics, knitting), is yes. Skill activities outlive the existence of the machines that perform them better, because the doing is the point.
Geography games might evolve. The format might become more about reasoning out loud, more about coached analysis, more about community play and tournaments. The core pleasure — looking at a piece of Earth and figuring out where you are — will not disappear. AI is reshaping every category it touches, but in this one, the audience seems to be growing not despite the AI but somehow because of it. The world is becoming more legible at every level except the one we look at with our own eyes, and people are quietly reasserting that level as a thing worth practicing.