"Geolocation" in OSINT specifically means determining where a piece of visual content was captured. The discipline has been refined over the past decade through the work of Bellingcat and similar investigative organisations and is now a teachable skill with documented methodology. The premise is simple: any photograph or video contains visual cues, buildings, vegetation, road markings, sun angles, shadows, signage, that constrain where on Earth it could have been taken. With enough cues and the right reference imagery, you can often pin a location to within metres.
The reason it matters: geolocation is the workhorse verification technique for any visual claim that includes a "this happened here" element. War-crimes investigations, disinformation analysis, missing-persons cases, real-estate fraud, and many other applications depend on it.
The basic workflow
A useful structure for any geolocation:
Start with what you can see. Inventory the distinctive features in the image. Buildings (architectural style, height, colour, materials, signage, decoration). Roads (markings, surface, direction-of-traffic, lane configuration). Vegetation (climate zone, ecotype, season). Vehicles (type, license plates, country-specific design cues). People (clothing styles, languages on signage). Sky and shadows (sun angle, cloud cover, time of day).
Estimate the country and region. Many features narrow location to a continent or country: language on signage, road markings, building style, vehicle types, plant species. The Geoguessr community has built extensive reference materials on country-specific cues.
Search for distinctive landmarks. Unique buildings, monuments, signs, or geographic features that can be searched directly. A McDonald’s logo with a specific design might be locatable by reverse image search; a distinctive piece of architecture might appear in tourist photographs.
Use mapping tools to verify. Google Earth, Mapillary, KartaView for street-level imagery. Sentinel Hub or Google Earth Engine for satellite imagery at various dates. The mapping tools let you compare candidate locations against the source image.
Cross-check with auxiliary information. Sun angle and shadow direction at a given date and time constrain location and time of day; the SunCalc website at suncalc.org is the standard reference. Weather records can confirm or rule out specific dates. Vegetation phenology (when leaves are out, when flowers bloom) can constrain time of year.
Document the chain of evidence. Screenshots of the source image, the matched reference imagery, and the reasoning. The verification trail is part of the deliverable.
The Bellingcat archive at bellingcat.com/category/resources/how-tos/ is the canonical public reference for the methodology. Individual case studies, the MH17 investigation, the Skripal poisoning, the Bucha massacre verification, show the technique applied to consequential investigations.
Specific tools
Mapping platforms:
Google Earth and Google Maps. The default. Excellent satellite imagery in most populated areas, street-view in many. The "historical imagery" feature in Google Earth Pro lets you compare imagery across years.
Mapillary (mapillary.com). Crowd-sourced street-level imagery. Often covers areas Google’s Street View does not. Owned by Meta but data accessible.
KartaView (kartaview.org). Open-data alternative to Street View; uneven coverage but useful where it exists.
Bing Maps. Sometimes has higher-resolution imagery for specific areas; "Bird’s Eye" oblique imagery is sometimes more useful than top-down.
Sentinel Hub (sentinel-hub.com). Free access to ESA Sentinel satellite imagery; useful for date-specific imagery and for finding visual changes over time.
Specialty:
Wikimapia (wikimapia.org). Crowd-sourced annotations on satellite imagery; sometimes has labels for features Google does not name.
OpenStreetMap and Overpass Turbo (overpass-turbo.eu). For querying OSM data, "find all cell-phone towers within 5km of this point," "find all buildings of type ‘industrial’ in this region."
Suncalc (suncalc.org) and Suncalc.net. For sun-angle and shadow calculation at specific times and dates.
Wolfram Alpha. Useful for time-zone conversions and astronomical calculations.
PeakVisor and similar mountain-identification apps. For locating images that include identifiable terrain.
A worked example
A simplified workflow for an image showing a specific street scene:
The image shows storefronts, a road, partial street signs in Cyrillic, and a building style typical of late-Soviet construction. First-pass: this is somewhere in the former Soviet Union.
The signage text, where readable, is in Russian, narrows to Russia, Belarus, parts of Ukraine, Kazakhstan, parts of Georgia, parts of the Baltics where Russian-language signage persists.
A specific shop name is visible. The shop name is searchable; the search reveals it is a chain with locations across multiple cities. The search returns a dozen candidate cities.
The architectural style includes specific details (decorative elements, building heights, balcony patterns) that match certain post-Soviet construction periods. Cross-referenced with city-specific architectural histories, the candidates narrow.
A road sign visible in the background includes a partial highway number. The numbering schemes are documented for each country; the partial number matches Russian highway numbering only.
Sun angle and shadow direction are estimated from visible cues. Combined with the date the image was published, SunCalc narrows the longitude band.
The combination of city-candidate list, longitude band, and the visible storefront shop reduces to two specific cities. Street-view exploration of likely streets in both cities, looking for matching architectural details, finds the location: a specific intersection in city X.
Documented confirmation: matching architectural details, matching road geometry, matching shop. Confidence level: high.
This is the kind of workflow that takes anywhere from minutes to days depending on the difficulty of the image. The Bellingcat investigations of specific war-crimes incidents have geolocations that took experienced investigators days of work each.
Common pitfalls
Several patterns produce false positives:
Confusing similar landscapes. A photograph of fields, sky, and trees might match many candidate locations. Without distinctive features, geolocation may not be possible.
Anachronistic reference imagery. Google Earth’s imagery may be years out of date for the area in question. A building visible in the source image but not the reference may have been demolished, or vice versa.
Manipulated source imagery. The source image itself may be a composite or manipulation. Forensic analysis is part of the workflow.
Cultural blind spots. Investigators who do not speak the language of the source area may misread signage, miss culturally specific cues, or mistake one country for another. Working with native speakers or language experts is sometimes essential.
Confidence inflation. The desire to produce a definitive answer can lead to over-claiming on weak evidence. Honest documentation of confidence levels is part of professional practice.
Geolocation as part of broader verification
A geolocated image is a building block in a larger verification. The geolocation tells you where; corroborating evidence tells you when, who, and what. The MH17 investigation by Bellingcat used geolocation as one of many techniques; the Skripal investigation similarly. The geolocation alone never closes a case; it is part of the case.
The 2026 state
The fundamentals have not changed. The tooling has continued to improve:
AI-assisted geolocation. Tools using machine-learning models to suggest candidate locations are emerging. Some commercial offerings exist; their accuracy is uneven and they cannot replace the manual verification step. They sometimes produce useful first-pass candidates.
Better satellite imagery. The combination of Sentinel-2, Maxar, Planet, and others gives higher resolution and faster update cadence than was available a decade ago. Some imagery is paywalled; Sentinel is free.
Increasing counter-OSINT awareness. Subjects sometimes deliberately strip metadata, blur backgrounds, or alter footage to defeat geolocation. The professional response is more careful collection and verification, not abandonment of the technique.
Greater public visibility. Geolocation has moved from niche skill to widely-taught technique through Bellingcat’s training programs, university courses on visual verification, and journalism schools. The community of practitioners is larger and more sophisticated than a decade ago.
For anyone interested in developing the skill: the GeoGuessr training games are an unexpectedly effective way to build the pattern-recognition muscle for country and regional cues. Bellingcat’s online courses provide structured training. The community on platforms like Twitter/X and Mastodon has weekly geolocation challenges that build practical experience. The skills compound; an investigator who has done a hundred geolocations is meaningfully better than one who has done ten.
The work is patient, methodical, and frequently consequential. It is one of the most accessible high-impact skills in modern investigative practice.