A new trend is gaining momentum online — users are actively using ChatGPT to determine the locations depicted in photos. After OpenAI launched its new AI models — o3 and o4-mini — the community was astonished by their capabilities: these models can meticulously analyze photos, even if they are blurry or cropped, and are able to find the tiniest clues to identify cities, venues, and even specific restaurants or bars.
The geoguessing power of o3 is a really good sample of its agentic abilities. Between its smart guessing and its ability to zoom into images, to do web searches, and read text, the results can be very freaky.
— Ethan Mollick (@emollick) April 17, 2025
I stripped location info from the photo & prompted “geoguess this” pic.twitter.com/KaQiXHUvYL
The X platform is flooded with examples where users upload restaurant menus, photos of facades, or random snapshots from walks into ChatGPT and ask the model to play “GeoGuessr” — a game where you have to guess the location from an image. The impressive results have sparked a wave of excitement, as o3 sometimes recognizes places in just seconds, even if the photo shows only a library interior or an unusual object in a bar.
Testing showed that even the previous GPT-4o model, which lacked advanced “visual” capabilities, often handled the tasks just as well as o3. However, o3 still managed to solve some images that GPT-4o could not — for example, identifying a famous Williamsburg venue by a purple rhinoceros head on the wall.
Alongside the excitement, concerns about privacy have also emerged in the community — the models can analyze photos without using metadata or chat history, which introduces new risks for users. OpenAI representatives stated that they have added refusal mechanisms for requests involving private information and are actively monitoring compliance with privacy policies, but existing cases show that security issues remain unresolved.
