Verifying viral video before and after publication

What a newsroom does when a viewer sends a video

A newsroom that asks its audience to send footage inherits a job it did not ask for: it becomes an investigation unit for somebody else's clip. You send a video, a phone number and a location, and someone on the desk now has to decide whether any of it is real. The work happens before a clip reaches the page and it keeps happening after, because a clip that spreads unverified does not stop at the first outlet that shares it.

The people who study this problem have a name for the confusion it grows out of. First Draft has spent years describing a polluted information environment, and its researchers separate the mess into categories rather than treating every bad clip as one thing. That distinction matters before you look at a single frame.

Disorder comes in more than one shape

A clip can be false, and it can be false in ways that change the response. Something can be invented outright. Something real can be moved out of its moment, a fire from one year presented as another. Something true can be framed to say more than it proves. A newsroom chasing all three with the same checklist will miss the differences, because the first needs a debunk, the second needs a date, and the third needs an argument about context.

When newsrooms learned to verify video as a routine, they borrowed habits from human rights investigators, who had already been doing this work with satellite imagery and open footage. The habits are unglamorous. Find the uploader. Ask what they saw and when. Look at the pixels, the shadows, the weather, the street furniture, and check the claim against all of it.

What first matters is patience with the obvious. A viral clip usually arrives already captioned, and the caption is often the part that is wrong. A newsroom that verifies the video but repeats the caption has done half a job. You can test this yourself the next time something arrives in your feed, by asking where the original was posted and whether the account posting it was there.

The clip travels faster than the check

Video spreads through group chats, short video apps and repost accounts, and each hop strips the source. By the time a local desk sees it, the original uploader is often three steps back and the timestamp is gone. A piece that looks impossible to find was posted days earlier with a location tag, and that tag survived on one mirror where nobody looked. Tracing a clip backward is slower than sharing it forward, and that gap is the whole problem.

Editors then face a decision with no clean answer. Publishing an unverified clip with a caveat still puts the images in front of people who will not read the caveat. Holding it lets a false version own the story for a day. Neither choice is free, and the second one gets worse when a competing outlet publishes first.

What a desk asks before a clip runs

The questions are short and the answers are hard. Who filmed it. When. Where. What is visible that could not be staged, and what is missing that should be visible. Whether anyone on the ground can confirm it independently. Whether the same clip has surfaced under two different captions, which is itself a finding. A newsroom that cannot answer most of these does not have a story, it has a lead.

Tools change, principles do not. Reverse image search on extracted frames. Map and satellite comparison for landmarks. Weather records for the hour claimed. The uploader's own history, which often settles it. The heavier work sits inside the story you already cover, whether that is Baltimore and Maryland coverage or a disaster that drew cameras from everywhere.

The infrastructure exists because the questions kept coming. First Draft collects its thinking on the contemporary information ecosystem, alongside research reports, training courses and classroom-ready guides. Its homepage groups that work by theme, including elections and public health, and the whole thing now sits hosted in perpetuity by the Internet Archive. What the page does not do is hand you a verdict on the clip in your inbox.

After publication, the desk keeps working

The clip that runs is not settled. A commenter spots a street sign that places the scene in another city. A resident says the building shown was demolished last spring. A competing newsroom finds the original, longer recording, and the shorter edit turns out to have removed the part that changed everything. Correction and amplification then compete for the same slot on the page.

Bigger desks own some of this infrastructure, and small ones do not. The correction is published, it is read by a fraction of the people who saw the clip, and the false version keeps circulating in the copy that was already downloaded. A newsroom that only plans for verification before publication has planned for about half the risk.

Why a correction rarely catches up

Once a clip is out, the correction has to compete with the original on reach, and the original almost always wins. People who saw the clip are not watching the corrections box. People who shared it are not notified. The false version is embedded in group chats and reposts that the newsroom cannot reach at all, and each repost lends its own small credibility.

This is where the frame of the story is decided, and where the habit of explaining the why pays off. If a desk says only that a clip is unverified, the audience learns nothing about how to check the next one. If it says which shadow gave the scene away, or which record contradicted the date, the audience can carry that test forward. Correction is not just retraction, it is teaching the room.

What training adds to the desk

A course on information disorder cannot make you a forensic analyst in an afternoon. What it can do is give a desk a repeatable set of questions, so that the check does not depend on one person's memory. The people building those courses argue that verification is a routine, not a heroic act, and routines survive staff turnover in a way that individual instincts do not.

The teams that handle this best tend to treat it like any other beat, with standing steps, a place to log what was checked, and a habit of saying publicly what could not be confirmed. That last part is uncomfortable. Readers often read a careful caveat as weakness, and an outlet that admits the gap may look slower than the one that just posts it.

Does an unverified clip belong on the page

Ask a room of editors and you get several answers. Show it with heavy framing, so the doubt travels with the images. Describe it in words, so the audience knows what is claimed without the footage doing the persuading. Or hold it entirely, on the theory that the clip will surface regardless and the newsroom's job is to be the one place that waited.

None of those choices is risk-free, and the trade-offs change with the story. In a breaking disaster, the gap between holding and publishing is measured in hours that people need. In a slow political story, it is measured in nothing, and patience costs little. A single rule across all of it tends to fail somewhere.

What is still not known

No one has a good account of how many false clips a typical desk sees in a week, because newsrooms do not publish their spike in submitted footage, and there is no shared count. First Draft's research reports studied the dynamics of a polluted information environment, and its training courses teach people to identify and respond to information disorder, but the numbers behind any one newsroom's inbox stay inside the newsroom.

So the next time a clip lands in your feed and asks to be believed, do the small part. Find the first account that posted it. Ask what date the scene could belong to. Look at one thing in the frame that a caption cannot fake, and start there.

About firstdraftnews.org

First Draft is a site that gathers thinking, research, training and visual resources on the contemporary information ecosystem. Its homepage groups that work by theme, including elections and public health, and offers guides, graphics and reports on information disorder. Its content is generally available under a Creative Commons Attribution 4.0 International license, and the site is hosted in perpetuity by the Internet Archive.