@coastline_nowPosted 2 min ago
Waves crashing over a seafront promenade under a huge storm cloud at dusk
“Storm hitting the coast right now”Capture data: none
AI traces
@coastline_nowPosted 2 min ago
“Storm hitting the coast right now”Capture data: none
What we readWhat the eye sees
Logged to the deskAnalysis kept, reason written
ConclusionSynthetic
Justifications
Generator signature found, AI-generated image
AI tracesHigh
No capture data

A viral picture,
and minutes to check it.

The photo desk that receives a dramatic image from a reader, a stringer or a social feed has to decide quickly whether it can run, and the storm, the crowd or the flood in it can now be produced by a model in seconds.

What it does

Checks the images, videos and audio that reach your desk, looks first for the signature left by the generator and then for the traces a model leaves in the content, and tells the editor what it found before anything is published.

Why it matters

A storm, a crowd or a flood can now be generated in seconds, and one published fake costs a newsroom far more than the minutes it takes to check.

Who it’s for

Newsrooms, photo agencies and platforms handling wire photos, reader submissions and social posts, and the verification teams who serve them.

The problem

Reverse search can only find a picture that existed before

Source checks and reverse image search were designed to catch a real photo taken somewhere else or at another time, and a picture generated a few minutes ago has no earlier version anywhere for them to find.

The breaking-news image

A storm, a fire or a protest that never happened, posted minutes after a real event starts and shared faster than anyone can check it.

The real photo with one detail changed

A genuine picture in which a sign, a flag or part of a crowd has been added or removed, so the scene is real and one element of it is invented.

City street after rain with a road closed barrierWhat your eyes see
What we see

A genuine street photo in which only the barrier and its reflection were added.

The series

Accounts that feed a story with dozens of images from the same model, each one plausible on its own and linked once you can read how they were produced.

In figures

Trust in what people see is already low

37%

of people trust most news most of the time, the lowest level since 2015.

Source: Reuters Institute, Digital News Report 2026
62%

of people say they are concerned about what is real and what is fake online.

Source: Reuters Institute, Digital News Report 2026
under1%

of news images and videos carry provenance data that could prove where they come from, by an estimate the institute quotes.

Source: Reuters Institute, Trends and Predictions 2026

Our answer

Where Label4.ai fits in your process

Label4.ai adds one question to the checks you already run, and leaves the rest of your process as it is.

Picture received

Wire, upload portal, e-mail or social feed, whatever your desk already uses.

Your existing checks

Source verification, reverse image search and the editor's eye. Unchanged.

Label4.ai analysis

Is this picture what it claims to be? Checked automatically as it arrives, in seconds, with its reasons.

Decision

Publish, publish with a label, or hold, with the reason written down.

A result anyone on the team can read

The editor sees the conclusion first, then three sentences they can repeat to the desk, like these.

  • A generator signature was found, which identifies this picture as produced by an AI model.
  • The traces left by the model cover the whole frame, so the scene itself was generated.
  • The picture carries no capture data from a camera.

How we do it

What sits behind each conclusion

  • An analysis shaped to your process Each project is set up for the way you work, with the checks that matter for your content, your threshold and the action we recommend. Every analysis looks first for a signature left by the generator, then for the traces a model leaves and for signs of editing.
  • Detection that keeps the process running Most pictures you receive are genuine, so before the contract we run your own content and agree with you on the false alarm rate that suits your process. That is why we quote no generic detection rate: we set up a project and tune it to give you the best detection rate on your own use case.
  • A conclusion that holds up The same content, analysed twice, gives the same report twice, which turns a series of pictures from suspicious into provable.

It runs in our cloud or on your own servers, through the Anomaly Engine console or the API.

As an option, with our watermarking expertise, the photos and videos you publish carry an invisible mark, so anyone can check that they really come from your newsroom. Anything without that proof is analysed as usual. About our watermarking

See the technology in detail

Your content

See what your own pictures reveal

You send a batch of pictures whose origin you already know, including the fakes you caught and the ones that ran. We come back with a result for each picture, your false alarm and missed detection rates, and a plain statement of what we did not see.

We reply to every request within two working days, under NDA if you prefer.