Claim MTR-2026-048213Attachment 2 of 4
Rear bumper of a grey hatchback, dented and scratched, photographed in a car park
IMG_4821.JPG, 11/03/2026, 23:14Capture data: edited
AI tracesManipulation traces
Claim MTR-2026-048213Attachment 2 of 4
IMG_4821.JPG, 11/03/2026, 23:14Capture data: edited
What we readWhat the eye sees
Logged to the claimAnalysis kept, reason written
ConclusionSuspect
Justifications
AI tracesHigh
Manipulation traces, bumper area
Capture data edited
No generator signature found

A photo in every claim,
and little time to look at it.

The claims handler who opens a motor claim is usually looking at a picture taken on a phone in a car park and sent through the app late in the evening, and what they have to settle is whether the payout can go ahead or whether the claim needs an expert, with a reason they can write down either way.

What it does

Checks every piece of a claim, from the photos and videos to the repair quotes and invoices, and tells the handler whether AI produced or edited it, in words they can paste into the claim.
See how we check documents →

Why it matters

A generated dent or a rewritten quote looks like any other piece of the claim, and the checks in place were never built to question how it was made.

Who it’s for

Motor, home, travel and gadget claims teams at insurers, MGAs and claims outsourcers, and the loss adjusters who work with them.

The problem

Claims checks were designed for reused photos and staged scenes

Reverse image search, claims rules and the handler's eye are good at spotting a picture taken from somewhere else or damage that was staged, and a photo of a real car with a dent generated into it gives them nothing to find.

The damage that was never there

A genuine photo of an intact car, with a dent generated into the bumper. The car, the plate and the car park are all real, which is what makes the picture so convincing.

The scratch that became a crushed panel

A small real mark, enlarged by a model until the repair estimate moves into another bracket. The claim itself is legitimate, and only its size has changed.

The quote with a new total

A real repair quote from a real garage, with the labour line and the total rewritten. The layout, the logo and the stamp are untouched, and only the edited lines carry the traces of the change.

The series

Three photos sent for one water damage claim, each from a different angle, and the same model produced all three. Each picture looks ordinary on its own, and the link appears once you can read how they were made.

Water damage claim, photo 1 of 3Photo 1 of 3
Water damage claim, photo 2 of 3Photo 2 of 3
Water damage claim, photo 3 of 3Photo 3 of 3

One water damage claim, three photos, all produced by the same model.

In figures

Claims fraud already approaches a billion euros a year in France alone

€947m

of insurance fraud was identified by French insurers in 2025, two thirds of it in motor, home and liability claims.

Source: ALFA, July 2026
99%

of claims professionals say they have already come across manipulated or AI-altered documentation.

Source: Verisk, March 2026
32%

of them feel very confident that they could recognise a deepfake.

Source: Verisk, March 2026
Projection for an insurer

Fraudulent claims stopped before payout, per year

1,000claims
1005,000
€3.7m

kept each year at €3,700 per stopped claim (AXA España, April 2026), before counting the hours your handlers no longer spend on photos that look fine.

Measure it on your own claims

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.

Claim received

App, e-mail, broker portal or phone line, whatever your intake already is.

Your existing checks

Reverse image search, claims rules, the fraud score you already run. Unchanged.

Label4.ai analysis

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

Decision

Settle, send an expert, or refer to the fraud team, with the reason written down.

A result anyone on the team can read

The handler sees the conclusion first, then three sentences written in the words of the job, like these.

  • This photo carries traces of a generative model, and they are strongest on the rear bumper.
  • The bumper area was modified after the picture was taken, which is where the damage appears.
  • No generator signature was found, so the conclusion rests on forensic analysis alone.

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 claims are genuine and should be paid quickly, 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 claims from suspicious into provable.

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

See the technology in detail

Your content

See what your own claims reveal

You send a batch of claim photos whose outcome you already know, including the frauds you caught and the ones you found out about too late. We come back with a result for each picture, your false alarm and missed fraud 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.