The Photo Types People Love Testing AI Geolocation With

There's a specific kind of fun in handing a machine a photo and watching it try. Not because you need the answer for anything — you're not lost, you're not trying to track anyone down — but because there's something genuinely satisfying about seeing an AI pick apart a scene the same way a sharp-eyed detective would: the roofline, the plug socket, the shape of the letters on a distant sign. If you've never tried it and aren't sure what to upload first, here's a rundown of the photo categories that tend to produce the most entertaining results.

The Mystery Childhood Photo

Almost everyone has one: a faded print from a family album, a house or a street corner from a trip decades ago, with nobody left to ask exactly where it was. These photos are a favorite for a reason — they usually have just enough going for them (a distinctive car model, a specific type of streetlamp, handwriting on a shopfront) that an AI guess can turn into a genuine "oh, that tracks" moment, filling in a small gap in a family story that nobody remembered to write down.

The 'Guess This Restaurant' Food Shot

A plate of food rarely gives away much on its own, but the frame around it often does — the tablecloth pattern, the style of chairs, a sliver of tiling or signage in the background, the particular kind of glass a drink was served in. People love running these through as a kind of game with friends: can the AI figure out which country a dish was eaten in using only what's visible around the plate? It's a nice reminder that geolocation clues are rarely the obvious ones.

The Car-Window Road Trip Shot

Snapped through glass, often slightly blurred, usually just a stretch of road, some fields or hills, and maybe a sign too far away to read clearly — these photos look like they'd be impossible to place, which is exactly why they're fun to try. Road markings, guardrail style, the color of soil, and the shape of distant vegetation can still add up to a surprisingly specific regional guess, even when there's nothing resembling a landmark in the frame.

The Generic Hotel Room Challenge

This one's become something of a running joke: hotel rooms are designed to look the same everywhere, on purpose, so a photo of one is treated as a deliberate stress test. People upload a shot of beige curtains and a nondescript bedspread specifically to see whether the AI throws up its hands or spots something — an outlet shape, a window latch, a view sliver through the curtains — that narrows it down anyway. Half the fun is watching how honestly the tool handles a scene that's built to give away nothing.

Other Favorites Worth Trying

• —Postcards and old prints. Scanned photos from a relative's shoebox, especially ones with visible film grain or a printed date stamp in the corner.
• —Screenshots from old home videos. A paused frame from a camcorder tape, often with the timestamp burned into the footage.
• —Airport and train station shots. Signage design and seating style vary more by country than people expect.
• —Backyard and balcony views. No landmark in sight, just plants, rooftops, and sky — a good test of how much climate and light alone can reveal.

Where to Try It

Raven, at withraven.net, is built exactly for this kind of casual experimenting — upload whatever you're curious about and get a best-guess location back, with nothing saved afterward; the photo is processed just long enough to generate an answer and then it's gone. If you'd rather do this on your phone in the moment a photo catches your eye, Geospy AI, our sibling app on the App Store, runs the same idea from your camera roll or camera directly, which makes it a natural fit for testing that mystery photo the second it comes up in conversation.

None of this is about proving anything or tracking anyone down — it's just a genuinely enjoyable way to look closer at photos you already have, and to be reminded how much information is quietly sitting in the background of an ordinary picture. Start with whatever photo you're already curious about; that's usually the best one to try first.