The principle
The geolocator matches features in the source footage — building shapes, road geometry, signage, terrain — to features visible in reference imagery (satellite views, street-view services, prior published photos). When enough features line up uniquely, the location is established.
Why it works
- Most places on Earth are visually distinctive at the meter scale.
- Reference imagery has become broadly available.
- Even adversarial actors usually post footage with enough visual detail to be located.
Why it sometimes fails
- Featureless terrain (desert, open sea).
- Nighttime footage with no visible landmarks.
- Heavily processed or low-resolution video.
- Locations where reference imagery is restricted.