Where was this taken? Start with what cannot move

Someone posts a photo with a caption: a flood here, a protest there, a delay on this road. The caption says where. The photo may say something different.

Working out where a photo was really taken is called geolocation. It means identifying the place a photo or video shows, usually down to a pair of map coordinates, by matching what is visible against maps and imagery. It matters because the answer to “where” often decides whether a claim holds up.

You do not need special tools to start. You need a habit of looking carefully and a method for deciding what to trust.

A four-step loop

The Kit, a free investigation guide from Exposing the Invisible, describes geolocation as a loop of four steps: collect information, process it, search, and check the result. If the check fails, you go round again with what you learned.

Here is how I would put each step to work.

1. Collect. List everything you can see, without judging it yet. Signs, shop names, road markings, rails, poles, building styles, a tower in the distance, the weather, the light. Include the context too: who posted it, when, and what they claimed. Write down what you cannot read or are unsure of, so you do not quietly turn a guess into a fact.

2. Process. Sort the list. Which clues are most distinctive, and which are most permanent? Then pick a starting point.

3. Search. Use your starting clue on a map or a search engine and collect candidate places.

4. Check. Compare each candidate with every clue you listed, not only the one you searched for. A candidate that fits one clue is a lead. A candidate that fits all of them is a result.

Start with what cannot move

The Kit gives a simple example. A shop name is a fixed element, because the shop will probably be in the same place next year. A car number plate is a movable one, because the car can be anywhere. Start your search with the fixed clue.

Ranking clues by permanence is the most useful single habit in geolocation. Roughly, from most to least reliable:

• Fixed: buildings, bridges, towers, rail or tram lines, road layouts.
• Semi-fixed: shop signs, bus stop signs, road signs. They can change, so check the date of anything you compare against.
• Movable: vehicles, people, parked objects, market stalls, temporary banners.

A clue that appears in only one place beats a clue that appears in many. Tram rails, for example, narrow a city to the streets that carry them. A pharmacy chain does not.

Chains, branches and look-alikes

Names are not unique. One of the worked examples in The Kit starts with a hospital name. The first match on the map was the wrong branch of the same hospital, and comparing street-level imagery with the photo showed the mismatch. Only then did the search move on to another branch.

Chains and franchises put the same sign in many places, so a matching name is a candidate, not an answer. This is the same lesson as with usernames: a match on a name is only as good as the evidence that it is the right one.

Left, right and which way you are facing

When you test a candidate on a map, track which side of the road things are on and which way the camera is pointing. A pharmacy on the left and a cafe on the right only match a map if you stand at the correct end of the street and face the correct way. Turn around and left and right swap.

Absence can be evidence

What is missing can rule a place out. Bellingcat’s early guide to geolocating videos gives an example: if the satellite map of a town shows no buildings taller than a particular mosque, but the footage shows one, that suggests the wrong location. In the same way, if a tower should be visible down a street but is not, that candidate is in trouble.

Be careful with this one. An item may be hidden by something else, or the imagery may be out of date.

Time and shadows

Shadows tell you when the picture was taken as well as where. A low sun gives long shadows. If you know roughly what time a photo was taken, which direction the shadows fall tells you which way the camera faces. Be careful where the time comes from: a posting time tells you when a photo was shared, not when it was taken. Bellingcat has written up a Shadow Finder tool that uses shadows to narrow down where and when an image was taken. You can start with the simpler habit of asking whether the shadows agree with the time and direction your candidate requires.

If the shadows disagree with the rest of your evidence, trust neither side yet. Check again.

Say how sure you are

A pin on a map is a claim, so attach a confidence level to it:

• High: several independent, permanent clues agree, and nothing contradicts them.
• Moderate: the main clues agree, but some depend on signs or imagery that may be dated.
• Low: you have one good clue and everything else is thin.

Map and satellite imagery has a capture date, and streets change. The Kit lists tools that show historical imagery, such as Google Earth Pro and Esri’s World Imagery Wayback, so you can check what a place looked like when the photo was taken.

Say what would raise your confidence, too: another photo from the same place, dated street-level images, or the poster’s other posts.

A word about privacy

Geolocation can expose where a person lives, works or goes to school. It is a good skill to learn and a bad one to aim at private individuals. Practise on material meant for practice, and when you do work on real events, ask whether the place you are naming could put someone at risk.

It also works the other way. Before you post a photo, look at it the way an investigator would: signs, distinctive buildings, a view from a window, even shadows can give away more than you meant to share.

Sources

• M_osint, Geolocation Methods: A step by step guide, The Kit, Exposing the Invisible.
• Bellingcat, A Beginner’s Guide to Geolocating Videos (2014).
• Bellingcat, Geolocation Techniques: Mapping Landmarks (2014).
• Bellingcat, Chasing Shadows: Geolocate Images with Bellingcat’s Shadow Finder Tool (2024).

Links accessed 12 October 2026.