Geospy AI on the Go: Photo Geolocation From Your Phone

Curiosity about where a photo was taken almost never arrives on a schedule. It hits mid-scroll through the camera roll on the sofa, in the middle of a group chat when somebody shares a mystery holiday shot, or while you are standing in front of an old print with no caption at a relative's house. In almost none of those moments is a laptop within reach. That is the gap Geospy AI was built to fill: the same visual guessing idea behind Raven, living where the photos already are.

Geospy AI is our free iOS app, and the experience is deliberately narrow. Take a photo or pick one from the camera roll, send it, and get back a best-guess location with a confidence read, produced by AI vision analysis of whatever is visible in the frame. There is no folder of saved results to manage and no fields to fill in. It is built around a single loop: photo in, guess out, purely for curiosity.

The narrowness is the point. An app that asked you to tag categories, add captions or curate a growing library of past results would work against the moment it exists to serve — a quick question in the middle of something else, not a filing project. So it stays out of its own way: point, guess, done, and you are back in whatever conversation or scroll you were in.

Standing in front of an unlabelled photograph at a relative's house, you can simply point the phone at it. Arguing over dinner about where someone's travel shot was taken, you can settle the matter on the spot rather than promising to check later. Scrolling your own camera roll and stumbling on a frame from years ago with no memory attached, you can have a guess in the time it takes to order another coffee. None of that works nearly as well when the tool lives in a browser tab you have to go and find.

This is where visual geolocation differs from a reverse image search. A search engine needs the picture, or something near enough to it, to exist somewhere in an index already; a photograph you took a moment ago exists nowhere. Computer vision approaches the problem from the other end, reasoning about the contents of the frame the way a well-travelled person would — a kerb profile here, a script on a shopfront there, a species of tree that only grows in certain latitudes. It also means the app never touches Exif, so a screenshot or a stripped file works as well as an original.

The same rules about photo choice apply on a phone as anywhere else, and they matter more than the device: a wide outdoor frame with context around the edges beats a beautiful close-up every time. The guide to the best and worst photos to upload is worth five minutes before you decide the model is guessing badly, and the quick-start guide covers the same ground in a shorter form.

Raven, at withraven.net, runs on Google's Gemini model and is built for sitting down with a larger screen: a stack of old family photos, scanned prints from a shoebox, a folder of travel shots you have been meaning to sort. The web version gives one free guess per account and then points to the free app. The commitment not to store what you upload is identical on both sides. Plenty of people use each in its place — the app for the in-the-moment "wait, where is this?", the site for an evening spent working through an archive. If you have never run the web flow, the step-by-step walkthrough covers it end to end.

• —Reach for the app when the photo is already on your phone, the question just came up in conversation, or you are away from a desk entirely.
• —Reach for Raven when you are going through a larger batch at once, working from a scanner or a bigger screen, or sharing the result with someone over a call.
• —Either way the privacy story is the same. Images are analysed for a single request and then discarded — never written to disk, never stored in a database, never used to track anyone.

Both sides of this rest on the same honest premise. A confident guess is still a guess, informed entirely by what is visible in the frame, and neither the app nor the website is trying to be anything more serious than a genuinely enjoyable way to satisfy a very old kind of curiosity: where, exactly, was this? Where that line sits, and why it is worth defending, is the subject of the piece on the ethics of AI photo analysis.