Methodology — Media Verification
Last updated: October 21, 2026
This is the order of work Citizen Analyst follows before it relies on, or debunks, an image or video in a published piece. It draws on the Verification Handbook, written by journalists from the BBC, Storyful and other newsrooms as a guide to verifying photos, videos and reports shared by the public in emergency coverage. The core rule here is to establish where content came from before deciding what it shows. The tools named here are described in Media Verification — Overview.
1. Claim decomposition
Split viral posts into atomic claims (who, where, when, what). Verify each independently.
A video can be authentic and still be misdescribed. “Real footage, wrong city” is a finding in its own right, and it is the usual shape of what Paris and Donovan call a cheap fake: real material re-contextualized. See Synthetic and Manipulated Media.
2. Source-first ordering
- Find earliest credible publication (reverse image / keyframe).
- Compare full context vs. crop.
- Only then use metadata or detectors as supporting signals.
For video, extract keyframes first. The InVID-WeVerify plugin splits clips from the major platforms into keyframes and sends them to reverse search on Google, Baidu or Yandex. Method: Reverse Image Search.
3. Multi-signal corroboration
Combine landmarks, language on signs, weather, shadows, and independent witnesses when available. No single signal is usually enough for high-stakes conclusions.
Shadows can be tested. SunCalc calculates the sun’s position, elevation and shadow length for any place and date, so a shadow’s direction can confirm or rule out a claimed time of day. Provenance data, where a file carries C2PA Content Credentials, is another independent signal.
4. Archival minimum
URL, capture time, tool, hash (when feasible). Note paywall or geo-block limitations.
Capture the original post before you publish anything about it, since the poster can delete it at any time. archive.today does not save video, so download the video file separately. Details: Web Archiving for Evidence.
5. Publication ethics
Label synthetic or uncertain clearly. Avoid identifying private individuals from images unless strong public-interest test is met.
State in the piece which checks were run and what each showed. A reader should be able to tell a confirmed location from a likely one.
Related skills
media-verification-specialist, geolocation-verification-specialist
Sources
- Verification Handbook (read 2026-10-21)
- Data & Society, Deepfakes and Cheap Fakes (Paris and Donovan) (read 2026-10-21)
- InVID project, InVID Verification Plugin (read 2026-10-21)
- SunCalc (read 2026-10-21)
- C2PA, Content Credentials (read 2026-10-21)
- archive.today, FAQ (read 2026-10-21)
Expanded with AI research and drafting assistance from the primary sources cited; editorial review pending.
