is this like GeoGuessr-solving AI?

Loosely. The vision pass reads the same kinds of clues a good player reads — architecture, vegetation, signage, light. But we pair it with hard EXIF GPS when that's present, and we're honest about confidence instead of always pinning something.

why do most of my photos have no GPS?

Social apps and messaging services routinely strip EXIF — GPS included — when you share a photo. Screenshots and downloaded images usually have none either. Photos straight off a phone camera are the ones that still carry coordinates.

does it work on screenshots?

Only through the vision path, since screenshots carry no GPS. It can still read the scene and make a region estimate, but expect lower confidence.

what file types do you accept?

JPEG and PNG are fully supported for EXIF; WebP, GIF and HEIC can be read for the vision pass. If a format confuses the parser, we just fall through to the model.

is my photo uploaded anywhere?

No original is ever stored. EXIF is parsed in your browser, and only a downscaled thumbnail is sent to our edge vision model for a stateless pass. See Privacy.

how accurate is it, really?

With GPS, it's as accurate as the coordinate embedded in the file — typically meters. Without GPS, it's a region-level estimate: country is often right, city is a coin-flip, and a street address is usually wrong. We tell you which one you're getting.

can I use this to find where a stranger lives?

No — and we'd ask you not to try. It's for your own photos and provenance checks. See the note on Privacy.