Screenshot Geolocation
Can you find where a screenshot was taken?
Screenshots have no EXIF, no GPS, no original metadata. They're often compressed on top of that. Orca reads the pixels — the buildings, roads, signage, and terrain — and geolocates the scene underneath.
How Orca reads pixels
Three classes of signal survive the screenshot pipeline. Orca is trained on all three.
Signal 01
Pixels
Orca reads the raw pixel grid: building geometry, road layout, terrain features, vegetation type. These survive screenshots and compression because they're structural, not metadata.
- ·Building proportions
- ·Road width and layout
- ·Terrain and topography
- ·Sky composition
Signal 02
Compression artifacts
Screenshots carry JPEG and platform compression signatures. Orca is trained on real-world imagery that has already been through social platforms, so it reads past the artifacts to the scene underneath.
- ·Platform compression
- ·Color quantization
- ·Resampling patterns
- ·Artifact-tolerant reading
Signal 03
Visual anchors
The strongest anchors are language-bearing signage, vehicle plates, utility pole design, and architectural vernacular. Each is a checkable fact that pins a screenshot to a place.
- ·Signage language and script
- ·Vehicle plate format
- ·Utility infrastructure
- ·Architectural style
Case study: verifying a screenshot from a scam chat
A buyer in an online marketplace was sent a screenshot of a "warehouse" to prove the seller had stock. The screenshot showed a loading dock, pallets, and a strip of daylight through a roller door. The seller claimed the warehouse was in Hamburg.
The buyer screenshotted the screenshot (already stripped of any metadata) and uploaded it to Oceanir.
Orca result
İzmir, Turkey
38.4237, 27.1428 · 61% confidence · Surface scan
Visual cues: Turkish-language safety signage on the dock wall, European pallet dimensions (EUR-EPAL), Mediterranean climate vegetation visible through the door. None consistent with Hamburg.
The buyer didn't send payment. The seller's "warehouse" was a borrowed photo from a logistics company in Turkey. The screenshot had passed through two messaging apps and lost every byte of metadata. The pixels were enough.
Why metadata fails and pixels don't
Every platform and every screenshot strips metadata. Visual geolocation sidesteps the entire problem.
Geolocate a screenshot
Free surface scan. No signup. Works on screenshots, forwards, and social exports.
Frequently asked questions
Can you geolocate a screenshot?+
Yes. Orca reads pixels, not metadata, so screenshots work. What you lose with a screenshot is EXIF, GPS, and file-level clues. What you keep is the visual scene: buildings, roads, terrain, signage, and vegetation. That's what Orca reads.
Do screenshots have any metadata left?+
Almost none. A screenshot is a fresh image of the screen buffer. It carries a new timestamp from the OS and a file format, but no camera EXIF, no GPS, and no original filename. The original photo's metadata is gone the moment the screenshot is taken.
How accurate is geolocation on a screenshot?+
It depends on the scene, not the format. A screenshot of an outdoor street scene with signage can resolve to street level. A screenshot of a face or an interior resolves to region or country at best. The confidence score on each result tells you how far to trust it.
What about screenshots from messaging apps?+
Messaging apps (WhatsApp, Telegram, Signal) apply their own compression on top of the screenshot. Orca is trained on platform-compressed imagery, so this is expected, not a blocker. Heavily cropped screenshots with no background reduce confidence significantly.
Can Orca read screenshots of maps or satellite views?+
Screenshots of map applications are a different problem. If the screenshot shows a recognizable street scene (Street View, photo mode), Orca can geolocate it. If it shows a top-down map tile, Orca may return low confidence because map tiles lack the physical-world cues it reads.
Is the surface scan free?+
Yes. The inline scan runs a free D1 surface analysis with no signup. You get coordinates, a confidence score, and a label. Deeper verification with evidence export and audit-grade PDF is available on a paid plan.