Claims and fraud teams
Damage photos, accident footage and travel images checked for place, date and editing before a claim is paid or refused.
OSINT for insurers →Geolocation Verification OSINT: Photos, Video and Media
Geolocation verification with OSINT tells you whether a photo or video shows what someone says it shows: the place, the date, the first upload and any sign of editing, each with the evidence behind it and a stated level of confidence.
- Geolocation
- Chronolocation
- Source and first upload
- Manipulation and deepfake checks
Short answer
In geolocation verification, OSINT analysts match what is visible in a photo or video, such as buildings, signs, terrain, vegetation and shadows, to satellite, street-level and other public imagery to establish where it was taken, then use sun position, weather and dated imagery to estimate when. OSINT-S adds source tracing and manipulation checks and grades every conclusion by confidence.
What Is Geolocation Verification in OSINT?
Four questions about one piece of media: where was it captured, when, who first published it, and has it been altered?
Geolocation fixes the place, ideally to a specific camera position. Chronolocation narrows the time, from a year down to an hour when shadows allow. Source tracing finds the earliest copy online and who posted it. Integrity checks look for edits, splices, recycled footage and synthetic content. A clip can pass one test and fail another: real footage from a real place, but filmed three years before the event it is shared as showing.
Metadata helps when it exists, but platforms commonly strip location and camera data on upload, and metadata is easy to edit. Our work relies on what is in the frame and in the public record, with metadata as a supporting check.
Who Uses Media Verification, and for What
Insurers, newsrooms, legal teams, NGOs and security teams, each with a different decision riding on the answer.
Newsrooms and fact-checkers
User-generated video verified before broadcast, with a note on what can and cannot be confirmed.
OSINT for media →Law firms and litigation teams
Photos and videos relied on by either side tested for authenticity, with a method an expert can explain.
OSINT for law firms →NGOs and human rights researchers
Documentation of abuses verified and preserved to standards that later investigators can build on.
OSINT for NGOs →Corporate security and crisis teams
Viral footage claimed to show your site, product or staff checked fast, before you respond publicly.
Investors and lenders
Site photos of a factory, mine or project compared with dated satellite imagery to see whether it exists as described.
How a Geolocation OSINT Team Verifies a Photo or Video
Preserve, trace the source, geolocate, chronolocate, test for manipulation, then grade confidence and send it for peer review.
- Preserve the originalObtain the highest-quality copy available, record where it came from, and hash it with SHA-256 so later versions can be compared.
- Trace the earliest sourceReverse image and keyframe searches across search engines and platforms to find the first upload and any older copies.
- Read the frameList every clue: skyline, road markings, signs and languages, shop names, vegetation, terrain, vehicle plates' format, weather.
- GeolocateNarrow the area, then match features against satellite, aerial and street-level imagery and mapping data until the camera position is fixed or the trail ends.
- ChronolocateEstimate time from shadow length and direction, compare with historical weather and dated imagery, and check for changes such as new buildings or roadworks.
- Test integrityLook for cloning, splicing, inconsistent lighting, re-encoding traces, recycled footage and signs of synthetic generation; check any Content Credentials.
- Grade and reviewAssign a confidence level to each conclusion; a second analyst repeats the key steps before a senior reviewer signs off.
Deepfake and Manipulation Checks
Detection tools give probabilities, not proof, so we combine them with source tracing and geolocation, which a forger finds much harder to fake consistently.
Altered images are already a claims problem. Allianz told UK media it saw a 300% rise in cases where apps were used to distort images, videos and documents between 2021-22 and 2022-23 (PA report, May 2024). Synthetic video adds to that.
Provenance standards help when present. C2PA Content Credentials bind signed statements about creation and edits to a file with cryptographic hashes, but the specification itself says it does not judge whether provenance is "good" or "bad", only whether the assertions validate (C2PA specification). A missing credential does not mean a file is fake, and a valid one does not prove the scene is true. That is why we test the content against the world: a shadow pointing the wrong way or a building that did not exist on that date is harder to explain away than a detector score.
Confidence Levels in OSINT Geolocation
Every conclusion is labeled, from confirmed to unverifiable, with the evidence that supports it and what would change it.
| Level | What it means | Typical basis |
|---|---|---|
| Confirmed | The location or date is established | Several independent features match reference imagery; time fits shadows and weather |
| High | Very likely, a minor gap remains | Strong match, one element not independently checked |
| Moderate | Plausible, alternatives not excluded | Area identified, exact position or date uncertain |
| Low | Possible, weak support | Generic features, few reference sources |
| Unverifiable | Cannot be determined from available sources | Close-up, indoor or featureless scene |
Some media cannot be placed: a close-up of a wall, a field at night. We say so rather than stretch. AI can help find candidate locations, but it is not a verdict. When Bellingcat tested 20 language models on 25 unpublished photos, the best only slightly beat Google Lens, all of them produced some entirely wrong answers, and the conclusion was that they cannot replace human geolocators (Bellingcat, 2025).
Berkeley Protocol and Evidence Standards
We follow the Berkeley Protocol's principles for collection, preservation and verification, and document methods so a court or editor can test them.
The Berkeley Protocol on Digital Open Source Investigations, published in 2022 by the UN Human Rights Office and the Human Rights Center at UC Berkeley School of Law, sets out how to identify, collect, preserve, verify and analyze digital open-source information professionally, legally and ethically. Written for international law, its disciplines carry over to commercial and civil cases.
Courts ask a simpler question. Under US Federal Rule of Evidence 901(a), the party offering an item must produce "evidence sufficient to support a finding that the item is what the proponent claims it is" (FRE 901). A verification report with preserved originals, reference imagery and step-by-step reasoning is built to help meet that test, whoever ends up presenting it.
What We Will Not Verify
We do not use geolocation to find private individuals, support targeting, or confirm the whereabouts of anyone without a lawful purpose.
- No locating people for personal reasons. Working out where an ex-partner or a stranger lives from their photos is stalking, not verification. We decline it.
- No targeting. We do not geolocate people, units or facilities to support attacks or military targeting.
- Protecting people in the frame. Bystanders, victims and witnesses visible in media are identified only when the case requires it, and graphic material is handled with care for the analysts and the people shown.
- Lawful collection. Media is gathered from public sources or supplied by you; we do not access private accounts to obtain it.
Turnaround and What You Receive
A short verdict first, then a full report with annotated images, reference sources, timeline and confidence levels.
A single image with clear features can sometimes be verified within a business day; a set of videos or a contested piece for litigation usually takes from 10 business days, and large collections up to about a month. Accelerated delivery adds 50%. You get annotated side-by-side comparisons, the earliest-source trail, hashes of the files examined and a methods note. Each report is peer-checked and signed off by a senior analyst, the quote is fixed after written scoping, and the fee goes down if we miss the agreed date. Verification often feeds a wider case, which is how it fits with our OSINT investigations and the rest of our OSINT services for legal, media and insurance teams.
OSINT Geolocation Verification: Specialized Services
Focused versions of geolocation verification for specific subjects, deals and situations.
Image and video verification
Whether a photo or video is the original, where and when it was taken, and whether it shows what the caption claims, each answer backed by evidence.
Read more →Deepfake detection
Suspect video, audio and images checked with detectors, provenance data and real-world context, reported with honest confidence levels.
Read more →Check It Before You Rely on It
Send the file or link, what it is claimed to show and the decision it supports. We tell you what can be verified and give you a fixed quote.
Geolocation Verification FAQ
A claimant sent us photos of storm damage to their roof — can geolocation verification with OSINT confirm the photos show their property and were taken after the storm date?
Often, yes. We match roofline, neighboring buildings, trees and street features to satellite and street-level imagery of the insured address, then compare shadows, weather records and dated aerial imagery with the claimed date. The report states a confidence level for location and timing separately. If photos are close-ups with no context, we may only be able to say they are consistent with, not confirmed as, the property.
Our newsroom received a video said to show a factory fire yesterday — how fast can you tell us whether it's really from that place and that day before we air it?
A first assessment of a single video can often be done within a business day: earliest upload, location match and whether the weather and light fit yesterday. You get a short verdict with a confidence level, then the full annotated report. If the clip turns out to be older footage, we show where it first appeared. We will not overstate a moderate match as confirmed to meet a deadline.
Opposing counsel produced a photo they say proves our client was at a meeting in another city — can you check whether it was altered or taken somewhere else, in a way an expert can defend?
Yes. We preserve and hash the file, trace any earlier copies online, geolocate the scene, check timing against shadows and weather, and test for edits and synthetic elements. Each step is documented with reference imagery so an expert can explain it and the other side can repeat it. Whether it is admissible is for the court; our job is a clear, testable method.
Can AI tools like ChatGPT do geolocation well enough now that we don't need a human analyst to check where our photos were taken?
Not on their own. AI models can suggest candidate places quickly, especially in cities with signs and distinctive buildings, and analysts use them as one input. But in Bellingcat's 2025 test of 20 models, all produced some entirely wrong answers, sometimes with confident explanations. For any decision that matters, a human has to confirm the match against reference imagery and record why.
We're an NGO documenting abuses from videos posted by witnesses — will your verification follow the Berkeley Protocol so our evidence can be used by investigators later?
We follow the Berkeley Protocol's principles: preserving originals with hashes and collection records, verifying source, location and time with documented methods, and protecting the safety of witnesses and people shown. We also flag what remains unverified. That gives later investigators a record they can test and build on, though acceptance by any court or mechanism remains their decision.
I found a photo of a woman on social media and want to work out exactly where she lives from the background — is that something your geolocation team will do?
No. Using geolocation to find a private person's home without a lawful purpose and her consent is the kind of tracking that enables stalking and harassment, and we decline it. Our work verifies media for insurers, newsrooms, legal teams, NGOs and companies, where the question is whether the content is authentic. If you are worried about your own exposure online, a digital footprint review is the right service.
Does a missing EXIF location or a stripped Content Credential mean that our insurance claim photos are probably fake, or can they still be verified?
Neither absence means fake. Most platforms and messaging apps strip metadata on upload, and Content Credentials are still uncommon. The C2PA specification itself does not judge whether content is good or bad. We verify from what is in the image, such as landmarks, shadows and dated satellite imagery, and treat metadata as supporting evidence when present, since it is also easy to edit.
Sources and Notes
- OHCHR and UC Berkeley: Berkeley Protocol on Digital Open Source Investigations (2022)
- C2PA Technical Specification 2.2 (Content Credentials)
- Federal Rules of Evidence, Rule 901 (Cornell LII)
- Bellingcat: Have LLMs finally mastered geolocation? (June 2025)
- Express & Star / PA: Insurers warn about fake and manipulated images being used in claims (May 2024)
Sources checked 7 October 2026. Figures about third-party firms and tools are as published by them or by the cited source on that date.