Guide Β· Media Literacy
A caption can put a photo anywhere it likes. Working out where the picture was actually taken is one of the most reliable ways to catch a mislabelled image β and you can get surprisingly far without any special software.
"This is happening right now in [city]" is one of the most common ways a real photo gets weaponised. The image is genuine; the place is a lie. Geolocation β figuring out the true spot a photo was taken β is how investigators test that. It's the same skill behind those viral threads that pin a photo to a single street corner, and the core of it is just patient looking.
Teams at Bellingcat and the Global Investigative Journalism Network geolocate images every day using free tools and a fixed order of checks. Here's that order, from the thirty-second version to the deep dig.
When a phone or camera saves a photo, it can tuck away hidden EXIF data: the device model, the exact timestamp, and β if location was switched on β GPS coordinates. When that data is intact, the whole puzzle collapses into one step. Drop the file into an EXIF viewer, and if coordinates are there, paste them straight into Google Maps.
The catch: it usually isn't there. Instagram, Facebook, X, and WhatsApp all strip EXIF the moment a photo is uploaded, so anything you pulled off social media has almost certainly had its coordinates wiped. Metadata is worth thirty seconds because when it survives it's decisive β but treat a blank result as normal, not suspicious, and move on to the picture itself.
Before you study the frame by hand, let the engines guess. Reverse image search will often recognise a famous landmark outright, or surface the same building from a differently-angled photo that does have a location attached. Crop tight on the single most distinctive thing in the shot β a monument, an unusual roofline, a stadium β and search that crop rather than the whole busy scene. Google Lens is best for well-known places; Yandex is stronger on ordinary streets and buildings. You're not trying to finish here, just to shrink the search from "the planet" to "this city" or "this coastline."
This is where most geolocation is actually won. Slow down and inventory everything the photo tells you about the world around it:
No single clue proves anything. The method is to stack them until only one region survives every constraint at once.
If there's a shadow in the frame, the sky is quietly telling you the time and latitude. The direction a shadow falls, and how long it is relative to the object casting it, both depend on where the sun sat β which in turn depends on location, date, and hour. Investigators call this chronolocation. Feed a candidate place and time into SunCalc and compare the sun position it predicts against the shadows in the photo. It's often less about pinning the exact spot than about a fast reality check: if a caption says "this morning" but the shadows point the wrong way for morning at that latitude, the caption is wrong. Contradictory shadows within one image β objects lit from different directions β are also a classic sign of a composite or AI-generated picture.
A guess isn't a location until you've matched it against reality. Take your best candidate and open Google Street View at that spot, then line up fixed features: the gap between two buildings, a lamppost, a particular window, the kerb line. Two or three independent features matching is a confirmation; one is a coincidence. Where Google's cars never drove, Mapillary's crowd-sourced imagery often fills the gap. This last step is what turns "probably somewhere in this city" into "this exact corner."
Upload a photo or paste a URL and it lays out every field the file carries β camera model, timestamp, and GPS coordinates if they survived. When coordinates are present it drops a pin on a map for you.
Strong at naming famous buildings and pulling up the same storefront or monument from other photos. Crop tight on one distinctive object β a sign, a tower, a bridge β and it works far better than feeding it the whole scene.
The best free engine for matching ordinary architecture and street scenes, and it indexes Eastern-European and Asian sources the others miss. When Google shrugs, this is the next stop.
Shows exactly where the sun sits in the sky for any place, date, and time. Match the shadow direction and length in your photo against it to test whether a claimed location and time are even physically possible.
Once you have a candidate location, ground-level imagery lets you line up the same kerb, lamppost, or window. Mapillary fills in streets Google's cars never reached, which matters outside big cities.
The photo came from Instagram, X, or WhatsApp
Assume the GPS data is gone. Almost every major platform strips EXIF on upload as a privacy measure, so a picture you saved from social media rarely carries coordinates. Skip the metadata step and go straight to the visual clues.
It's an indoor shot with no windows
No sky means no sun, and often no recognisable exterior. Hunt for text instead β a menu, a certificate on the wall, a price tag, a plug socket type, a fire-exit sign in a particular language. One readable brand or phone number can be searched on its own.
The landscape is generic β a field, a beach, a forest
Don't expect a pin. Stack several weak signals: vegetation type, the angle and length of shadows for a latitude estimate, any distant infrastructure (pylons, a road, a fence style). Weak clues combined can still rule whole continents in or out.
Nothing lines up anywhere
A picture that resists every check may be composited or AI-generated rather than a real place. If the scene has that too-smooth, plausible-but-placeless look β and shadows that point in contradictory directions β stop geolocating and switch to AI-image checks.
A wave of AI "photo location" tools now guess a place from visual content alone, and they're genuinely good at naming a country or a rough region in seconds. Treat that as a starting point, not an answer: a model's confident guess is a lead you still have to confirm on Street View before you rely on it. The reverse is worth remembering too β the same techniques that locate a public building can expose where a private person lives. Use geolocation to test a claim about a photo, not to track someone who never asked to be found.
If a photo comes attached to a viral story, the location is only half the question. FAXTR searches 100+ fact-checking organisations across 11 languages in one box β free, no login β so you can see whether the claim already has a published verdict.
Go to the verifier β