Mergen Demo Investigative 13/60
Skip to content
Genel

How do fake images spread? Twenty-four hours with a fact-checking team

We spent a day with a fact-checking team that tracks down fake images spreading on social media during crises. In our video, they explain their methods and the challenges they face.

JMJake MoralesSports & Technology Reporter
7 min12.3B2
YouTubeYouTube content loads when you press play
Table of contents
  1. Tracing an image
  2. Reverse image search and metadata
  3. AI-generated content
  4. Key takeaways

After a major natural disaster, a significant share of the images shared on social media turn out to be photos taken years earlier in another country. In our video report, we spent twenty-four hours with an independent fact-checking team and follow, step by step, how editors trace the source of an image and use reverse image search and metadata analysis. The team says AI-generated content is making the job harder every day.

Tracing an image

The fact-checking team's day starts at 5 a.m., scanning the previous night's crisis posts. Editors quickly sift and prioritize hundreds of posts: which ones were shared widely, which involve casualties or institutions, which look suspicious. Then a file is opened for each image, and the first question is asked: who first published this image, and where? The answer often goes back much further than expected.

The team's rule is never to comment on an image before verifying it. That can sometimes take hours. Meanwhile, the fake image keeps circulating fast. The team leader says speed never comes before accuracy, but silence isn't the answer either. So until an image is verified, viewers are shown a "not yet verified" note. The editors believe this is the most honest way to protect their audience's trust.

A lie can travel halfway around the world while the truth is still putting on its shoes.

Jake Morales

Reverse image search and metadata

A reverse image search debunks most fake images within minutes: the same frame turns out to have been shot years earlier in another country. But when nothing turns up, the work falls to metadata and contextual clues. Small details are examined one by one, from file properties and capture dates to the writing on signs, the direction of shadows and license plates. Every clue is logged in a separate notebook and rechecked by a second editor.

One of the examples the team spent the most time on in the video is a frame that appears to have been shot in the rain but was actually taken from a film set. What gave it away was the reflection of lighting equipment in the background. One of the editors explains that the method isn't magic; it comes down to the discipline of patient comparison and careful note-taking. In her view, the biggest mistake is to rely too heavily on a single finding and ignore the rest.

AI-generated content

AI-generated images are complicating the fact-checking desk's work. It used to be enough to find a source photo; now there are convincing frames that have no source at all. With this kind of content, the team looks for inconsistencies such as the number of fingers, the shapes of text and reflections, but says these can no longer be treated as conclusive proof. They add that as generation tools improve, these telltale signs may soon disappear too.

Topic Analysis
Highlights disinformation, investigation, artificial intelligence
Reading time 7 minutes
Written by Jake Morales

Key takeaways

  • Verification starts with finding out who first published an image, and where.
  • Reverse search and metadata quickly debunk most fake images.
  • With AI-generated content, relying on a single clue isn't enough.

That is why the team never relies on a single clue and focuses instead on the source's history and the moment the content first spread. They share one simple rule for everyone watching: before you share an image that triggers an emotional reaction, ask where it came from. At the end of our video, we also walk through the team's checklist step by step. It includes simple steps anyone can do at home, such as checking the source, date, location and context.

Share this post
JM
Sports & Technology Reporter

Jake Morales

With field experience ranging from Premier Division stadiums to volleyball arenas, Jake Morales follows cybersecurity and artificial intelligence as well as sports news. He is part of the team that tracks the live news feed on publication days.

2 Comments

Share your thoughts

Your email address will not be published. Respectful, on-topic comments are approved.

News Central

Welcome back

Log in to your account; your saved posts and comments are waiting for you.