A photo or a video often arrives with a claim attached: this was taken here, on this day. Image geolocation is the discipline of testing that claim by working out, from what is visible in the frame, where the picture was taken. Journalists, fact-checkers and investigators use it to confirm events, expose recycled images and place a scene on a map. This guide covers the basics: where to look, how to narrow a search and how to avoid fooling yourself.
Start with the question and the original file
Write down the claim you are testing and what you already know: who posted the image, when it first appeared, and what it is said to show. Work from the largest, least compressed copy you can find, and keep the original untouched. Zoom into each part of the frame in turn. Small details such as a sign, a shop front or a utility pole often carry more information than the main subject.
Metadata: useful, but easy to lose
Photos taken on phones and cameras can carry EXIF metadata, including the capture time and sometimes GPS coordinates. Treat it as a lead and not as proof. Most social networks and messaging apps remove metadata when an image is uploaded, so its absence says nothing. When it is present, it can be edited, and the camera's clock may have been wrong. Check it against what you can see.
Visual clues to look for
- Language and scripts. Shop names, street signs, advertising and the way phone numbers are written narrow the country or region quickly.
- Roads. Driving side, lane markings, the shape and color of road signs, bollards and guardrails differ between countries.
- Utilities and street furniture. Utility poles, street lamps, bus stops and mailboxes follow local designs.
- Architecture. Building materials, roof shapes, balconies and window styles suggest a region and an era.
- Landscape. Terrain, vegetation, soil color and the visible skyline point to climate and geography.
- Vehicles. License plate formats and taxi and bus designs are regional.
- Landmarks. Towers, bridges, stadiums and religious buildings can anchor a search.
- Sun and shadows. The direction and length of shadows give a rough idea of orientation and time of day, which can also test a claimed time.
Narrow it down like a funnel
Move from the large to the small. Use broad clues to settle on a country, then regional clues to pick a region, then local clues to find a town and a street. At each level, form a hypothesis and look for something that would disprove it. Once you have a candidate area, compare it with maps, satellite imagery and street-level photos, and match several details at once: the road layout, the skyline, the shapes of buildings, the position of trees. One matching detail is a coincidence. A cluster of matches is a location.
A reverse image search can also help. It can show whether the image appeared earlier, which would contradict a claimed date, and it can lead to other pictures of the same place.
Verify before you conclude
- Look for a second source. Other photos or videos from the same spot, local reporting or an independent angle.
- Check consistency. Season, weather and vegetation should fit the claimed date. Look for signs of editing or compositing.
- Be wary of your own expectations. Confirmation bias is the biggest risk. Write down what would prove the location wrong, then try to find it.
- Expect fakes. Staged scenes and AI-generated images are part of the landscape.
A worked example
A photo circulating online is said to show a street scene in one city, taken yesterday. A first pass could look like this:
- Claim: taken in city A, yesterday.
- Origin: a reverse image search finds the same photo posted three years ago. That already contradicts "yesterday".
- Language: the shop signs use a script and a language of city B's country, not city A's.
- Roads and street furniture: traffic drives on the side used in country B, and the bollards and the bus stop match local photos from there.
- Landmark: a distinctive tower in the background matches street-level imagery of one district of city B.
- Conclusion: the photo is old and was taken in city B. The claim does not hold, and the matching details and the earlier post are the evidence.
Where AI geolocation fits
AI geolocation tools estimate where a photo was taken from learned visual patterns. They are most useful as a source of hypotheses: a region or a few candidate places to test against maps and imagery. Their answers are probabilities, not certainties, and they struggle with indoor scenes and generic landscapes that have no distinctive features. Treat the output like any other lead and verify it with the methods above.
Ethics and safety
Geolocating a picture of a private home, a child or a vulnerable person can put someone in danger. Verify before you publish, and follow the privacy laws that apply to you. Harassment, stalking and doxxing are never legitimate uses. Our Ethical Guidelines describe what we do not allow. This is general information, not legal advice.
Where OSINT Search Engine fits
OSINT Search Engine includes an AI image geolocation search that returns a probabilistic location estimate for a single scene or place photo, alongside its email, phone, username and domain searches. Treat the estimate as a hypothesis to test. See the OSINT Search page, and the guide What Is OSINT? for the verification habits behind it.
