VERIFICATION · JOURNALISM & OSINT
Verify a photo's location
A claimed location is a testable claim. Here is the evidence workflow for testing it — and a free tool that runs the metadata and clue stages with its reasoning on show.
The most common way a false photo spreads is not a fake image — it is a real image with a false caption. A genuine photograph of one city, one protest, one flood, relabelled as another, will outrun any correction. Checking a claimed location is how that gets caught: not by asking whether the photo “looks right”, but by testing whether the evidence in and around the file is consistent with the place it claims to be.
Location verification is a consistency exercise. A photo can be consistent with its claimed location, inconsistent with it, or — very often — simply unverifiable from what is visible. Each of those is a legitimate finding, and confusing the third for either of the first two is how verification itself goes wrong. The workflow below is the one open-source investigators use, and the free tool on this page automates the metadata and clue stages of it — with its reasoning on display, and its limits stated in the open.
Those limits, up front: the tool returns areas, not pinpoint addresses — never tighter than a 250 m radius. It works on public places, and suppresses results for homes and private property to a general area. It does not run reverse image searches and does not compare photos against street-level imagery. It is one stage of a verification, not a verdict — and it is not a surveillance tool: it analyses a photo you provide, identifies no people, and performs no face recognition.
What verification can — and cannot — establish
Location verification answers a narrow question: is the visible and embedded evidence consistent with this photo having been taken at the claimed place? It cannot establish when the photo was taken (an old photo of the right place is the classic recycling trick), who took it, or whether the event it supposedly documents happened as described. Those are separate checks, with separate evidence.
Keep the three outcomes distinct in your notes and in your copy. Consistent: independent clues support the claimed location and none contradict it. Inconsistent: at least one strong clue contradicts the claim — the photo is somewhere else, whatever its caption says. Unverifiable: the file and the frame between them do not carry enough location evidence to test the claim at all. Most viral imagery, handled honestly, lands in the third bucket — and saying so is a service, not a shrug.
The evidence workflow
1. Preserve the original. Every hop through a messaging or social app strips metadata. The file your source's phone produced is evidence; the screenshot of it in a group chat is a picture of evidence. Ask for the original, as a file, before you analyse anything.
2. Read the metadata as a claim. EXIF GPS and capture time are the first things to check — and the first things a motivated faker edits. Note what the file asserts, then hold it to the same standard as the caption: it has to survive cross-examination by the scene.
3. Extract the scene's clues. Signage and its language and script, named landmarks, licence-plate formats, driving side, road markings, architecture, vegetation. This is the stage the tool automates: its visual pass extracts clues only and never names the answer itself, so the ranking that follows is a deterministic product of the evidence, not a model's hunch.
4. Match against map data. Named places are looked up in OpenStreetMap and every candidate is scored on how much independent evidence supports it — and what speaks against it. Both columns are shown. In verification, the “against” column is the one that keeps you honest.
5. Check the place itself. A result on WorldEye links onward to live cameras near the candidate location and an Explore location in 3D view on the globe. Cameras show the place as it is now, not as it was when the photo was taken — but terrain, street layout and permanent landmarks change slowly, and a scene that cannot exist at the claimed location is a decisive contradiction.
Run the check on a photo
Upload the best copy you have. You get the metadata findings, ranked candidate locations with the evidence for and against each, a confidence level and an uncertainty radius — or an honest “couldn't determine”. Free beta, no signup, and the photo is processed in memory and never stored.
Open the Photo Geolocation toolFree beta · 3 analyses a day · Evidence shown for every candidate
How to check a claimed location
- 1Preserve the best copy you can get. Ask the source for the original file: screenshots and app-forwarded copies have had their metadata stripped before you ever see them.
- 2Form the claim precisely. “This photo shows place X” is testable; note who is claiming it and what would follow if it were wrong.
- 3Run the photo through the Photo Geolocation tool. It reads any EXIF GPS in the file and extracts the scene's visible clues — signage, landmarks, plates, driving side, scene type.
- 4Compare the ranked candidates with the claim. A candidate that matches the claimed place, backed by independent clues, supports it. Note the confidence level and radius — a broad area match is weaker support than a tight one.
- 5Hunt for contradictions. One strong inconsistency — traffic driving on the other side, signage in the wrong language or script, a named landmark that is somewhere else — can outweigh several weak agreements.
- 6Corroborate before you conclude. Cross-check with live cameras near the candidate on WorldEye, contemporaneous reporting, and other primary sources. If the evidence stays thin, publish the honest finding: unverified.
Limits, stated plainly
This is a beta, and verification is exactly the field where overstated tools do damage, so here is the whole list. Results are areas with an uncertainty radius, never displayed tighter than 250 m. Coverage is public places: results for homes and private property are suppressed to a general area, always. There is no reverse image search and no matching against street-level imagery, so the tool cannot find earlier copies of a photo or confirm a scene against ground-level reference photos. The visual pass reads clues; the ranking is deterministic — the same evidence always produces the same order — but clue extraction can miss things a human would catch, and can misread. When the daily visual-analysis budget is used up, the tool continues in metadata-only mode and tells you so.
And a boundary that is not about accuracy: this tool analyses imagery of places. It does not identify people, performs no face recognition, and must not be used to locate private individuals. Verifying a public claim about a public place is journalism; tracking a person is not, and the area-level, public-places design is what keeps the tool on the right side of that line. For the wider craft, see OSINT for journalists and our walkthrough of verifying breaking-news footage. If your need is the everyday version of this question — your own photos, your own trips — the Photo Location Finder is the same tool framed for travellers, and where was this photo taken explains the metadata side from the beginning.