Verifying a Viral Image

This case study is an educational composite that draws on the public methodology practices of Bellingcat, First Draft News, and the Reuters and AP verification teams. Names and specific images have been anonymised. All techniques described are for lawful, ethical use.

Context

A photograph circulated on social media in late 2024 showing what appeared to be a protest in a major city, captioned with a date in the preceding week and a political claim. Within twelve hours the image had been reposted tens of thousands of times, quoted by partisan outlets, and shared by accounts with large followings. A newsroom verification desk picked it up for a brief.

This write-up reconstructs the steps a competent verifier would take against a case like this. The specific image is fictional; the methodology mirrors documented public work.

Question

Is the photograph what the caption claims — a protest in that city, on that day, about that subject?

Verifiable subquestions:

1. When was this image first published online?
2. Where was the photograph taken?
3. Does any available metadata contradict the caption?
4. Who is the earliest identifiable publisher?

Methodology

Planning. The desk set a one-hour budget for an initial triage and agreed explicit acceptance criteria: the image would be published only if the capture location, capture date, and subject could each be independently corroborated.

Collection.

1. The highest-resolution available copy was saved from the originating post. The post URL and the user profile URL were preserved via archive.today and web.archive.org/save/ before further work, following the Wayback Machine workflow.
2. ExifTool was run against the saved file. EXIF data had been stripped — consistent with upload through a major social platform, which is itself a lead, not a conclusion.
3. The image was submitted in parallel to Google Lens, Yandex Images, Bing Visual Search, and TinEye, per the reverse image search tutorial.

Analysis. TinEye's oldest-match sort returned a copy published 19 months earlier on a regional news site, with a different caption describing a different event in a different city. Yandex surfaced two additional earlier copies on non-English outlets, consistent with the TinEye date. Google Lens surfaced only later re-uses.

The earlier publication included a named photographer byline. The verifier contacted the news desk that held the original and was sent a copy with intact EXIF. Camera model, timestamp, and GPS matched the regional-outlet caption. The GPS coordinates placed the photograph 2,100 kilometres away from the city the viral caption claimed.

Tools used

• Reverse image search for the initial identification of earlier publication.
• Metadata extraction on both the viral copy and the source copy.
• Wayback Machine for preserving the viral post before any public pushback.
• Google dorking to locate re-uses of the image and distinctive caption fragments in quoted form.

Evidence snapshot

Findings

1. The viral image predates the claimed event by more than a year.
2. It depicts a different city, on a different subject, by an identifiable photographer.
3. The viral caption is either a fabrication or a misattribution, depending on whether the original poster knew of the earlier publication.
4. Several downstream accounts with large followings republished the image without verification and continued to do so after public corrections appeared.

A retraction was issued by the reposting outlet within 36 hours of the verification desk's brief. The originating viral post remained online but received an in-platform context label.

Lessons learned

• Archive first, analyse second. The original post was edited by the user within 24 hours; the archived snapshot preserved the original caption and timestamp.
• One-engine reverse searches miss cases. The decisive earlier-match evidence came from Yandex and TinEye, not Google Lens.
• Direct contact with the original photographer is both the strongest corroborator and the most often skipped step.
• Platform-stripped metadata is not the absence of metadata — the stripping itself is a fingerprint, and the pre-upload file usually survives somewhere.
• A documented chain of custody, even for a brief verification, is what allows the finding to survive pushback.

Ethical considerations

For continuing work on the methodology behind image-led investigation, see the Epstein Revealed investigation series and the foundational writeups at Bellingcat. For the rights framework when civic documentation is itself part of the story, see the ICE Encounter rights guides.