HN Explorer

Show HN: Daily image location guessing agains AI (mediaguesser.com)

1 points by jonamedia ↗ HN
Every day, I let all frontier AI models play five rounds of an image location guess game. Try it out how you perform against the AI.

I recently released my browser image location game, a hobby project where you’re get random images from sources like Wikimedia Commons and have to guess the image’s location. There are an incredible number of images with rich metadata on Wikimedia Commons, most of them depict trivial but interesting places or events. My goal was to make this online catalog playable and to make it easier to discover new places and Wikipedia articles. I realize I’m certainly not the first person to have this idea, but I’ve tried to build a cohesive overall concept around it with multiplayer, content filter, own image collections and more.

I thought it would be interesting to see how well AI models do at guessing images, and I have to say, they're very good. To make it a little more interesting, every user can also take a guess at each image. I hope you find it interesting and enjoy guessing. I know it's not really a fair comparison. In the end it's more entertainment then a real competition.

Please feel free to ask any questions and I look forward to any suggestions/feedback.

3 comments

[ 0.22 ms ] story [ 29.6 ms ] thread
tilemarch
Do you know how the AI agents do it ? are they cheating by indexing the sources of your images? :)
jonamedia
Interesting question but I don't think they're cheating. Each model only receives the bare image (re-encoded, hosted under a random ID and with the EXIF data stripped), plus the question: Where was this photo taken?. We don't provide them with web search or any other tools. Looking through past rounds, nearly all their near-exact guesses are photos with readable text (e.g. signs, place names or plaques) or landmarks. In contrast with more generic photos they're typically 50–400 km off. Memorising the training data is a fair assumption, since all the photos come from public sources. However if that were the case, they would probably guess the obscure photos just as precisely, but the current data suggests that they do not. One photo taken in mid-2026, after most of the models had been trained, was still placed within 3km. My guess is during training, they are likely to see huge numbers of images with locations, so they are very good at recognising places and reading signs in any language.
dongjoo
It’s amazing how AI models can pinpoint such precise locations, even on ordinary-looking streets. The page design is also clean and attractive. It would be fun if there were a hint feature, too.