Before You Share: Verifying a Viral Disaster Video
Test whether footage really shows the place and event in its caption, using keyframes, maps, weather records and independent reporting.
Illustrative scenario: a viral clip claims to show today's flooding in a named city. The useful question is whether that caption is supported, not whether the uploader seems believable. Separate the claim into location, time and event. Footage can be genuine while its caption is wrong.
Preserve the claim and search for earlier footage
Record the public post URL, the caption and the time you saw it, including the time zone. Extract several distinctive frames with a video-verification tool and run reverse image searches on the frames and selected crops. A matching earlier upload may disprove 'today,' but its caption is another claim to verify. The earliest copy you find is not necessarily the original.
Match fixed features, not the floodwater
- List visible street names, bridge shapes, shopfronts and road junctions. Avoid relying on colours or weather alone.
- Use maps and dated street-level imagery to test candidate locations. Compare several fixed features and their relative positions.
- Check whether apparent landmarks existed at the claimed date; map imagery can be years older than the post.
- Keep competing candidates until the evidence rules them out. A familiar-looking bridge is not enough to identify a city.
Test the time independently
Compare local weather observations, official flood notices and independently gathered local reporting. Explain the limits: a station outside the city may not capture local rainfall, and a weather match cannot date a video by itself. The upload timestamp establishes when the platform received that copy, not when filming occurred. Multiple outlets embedding the same clip are not multiple independent observations.
Publish the smallest defensible claim
An honest outcome could be 'the bridge matches this location, but the filming date remains unconfirmed.' Link the original sources and state what would resolve the gap. Do not name bystanders, expose a shelter's sensitive location or reuse victims' images unnecessarily. Visual oddities and automated AI-detector scores are not enough to declare footage fabricated. If time and place cannot be established, say so rather than passing the caption along as fact.