GeoFinderAI

About GeoFinderAI

GeoFinderAI predicts where a photograph was taken by reading the visual evidence in it, which is the task human geolocation hobbyists do manually and which turns out to suit a vision model well. It interprets architecture, signage and script, terrain, sky conditions, vegetation and cultural markers, builds a location hypothesis from them, and returns coordinates with a confidence percentage plus a region, country and city breakdown. Predictions typically complete in under thirty seconds, and the model is trained on geotagged imagery spanning more than 190 countries, covering rural landscapes as well as dense urban centres. The site reports over 1.2 million images analysed and more than 10,000 users, and its second-generation model is described as 43 percent more accurate than the first. The analysis is presented in stages with similarity scores shown rather than as a single opaque answer, which matters for the OSINT and journalism use cases where a claim about location has to be defensible rather than merely plausible. Three free predictions run without a credit card. Uploads are handled privately. Pricing is one-time token packs rather than a subscription, at one cent per token with ten tokens per location guess.

Key Features

Tags