Product · for claims & compliance teams

Image & Document AI

Automatic screening of photos and documents — metadata forensics, tamper detection and cross-checks against the weather that really happened. Honest files sail through; suspicious ones get flagged with reasons.

Image & Document AI forensic dashboard: tamper, splicing and clone detection on a storm-damage photo, weather cross-check for the claimed time and place, invoice analysis and an AI authenticity verdict.
Metadata forensics Tamper detection Weather cross-check API for your claims flow
What it does

Three layers of checks on every file

Each photo and document is screened for what it says about itself, whether it has been altered — and whether it matches reality at the claimed time and place.

Metadata forensics

Capture time, device, GPS and edit history read from every file and compared with the claim.

Tamper detection

Signs of editing, splicing or AI-generated imagery flagged with confidence scores.

Weather cross-check

Does the damage match the weather that actually occurred there? Our 20-year archive answers.

Automated validation

Your rules turn scores into statuses — pass, review or reject — inside the claims flow.

Review queue with reasons

Flagged files arrive with human-readable explanations, not black-box scores.

Batch & API

Screen single files or whole backlogs; results as JSON with full detail.

How it works

From upload to verdict

STEP 01

Submit the files

Photos and documents arrive via API straight from your claims system — or in bulk for a backlog audit.

STEP 02

AI screens every layer

Metadata, manipulation signals and the weather cross-check run automatically on each file.

STEP 03

Clear — or flagged with reasons

Clean files pass silently; suspicious ones land in a review queue with the evidence attached.

Two ways to run it

Review dashboard for people, API for the flow

Review dashboard

For investigators & claims teams

A queue of flagged files with side-by-side evidence — built for fast decisions.

  • Flags with plain-language reasons
  • Weather at the claimed time and place, inline
  • Case notes and export for the file

Screening API

For claims systems

JSON in, verdict out — wired into intake so screening happens before an adjuster ever looks.

  • Single-file and batch endpoints
  • Scores, reasons and statuses in one response
  • Webhooks for review-queue events
At a glance
Inputs
JPEG, PNG, HEIC photos; PDF and common office documents.
Checks
Metadata consistency, manipulation and AI-generation signals, weather cross-check, document integrity.
Output
Scores with reasons and a status (pass / review / reject) as JSON.
Weather basis
20 years of hourly archive from a dense station network — official and our own — radar and our models.
Data handling
Files processed for screening only; retention and hosting per agreement.
Delivery
REST API, webhooks and a review dashboard.
Questions we get

Before your fraud team asks

How accurate is it?

Every flag comes with a confidence score and reasons, and thresholds are tunable — you decide what auto-passes and what a human sees. A pilot on your closed cases shows the real numbers.

Can it catch AI-generated images?

We flag statistical signs of generated or heavily edited imagery. It’s an arms race, which is why we layer it with metadata and the weather cross-check — context is much harder to fake than pixels.

What if metadata is missing?

Missing or stripped metadata is itself a signal, and the weather cross-check still works from the claimed time and place.

How does it fit our claims system?

Screening runs at intake via API, so results are attached before triage. The dashboard is optional for the review queue.

Why it works
“Fraud rarely survives context. A photo carries its own testimony — when it was taken, where, and whether the sky agrees with the story.”
Forecasts.cloud · forensics with a weather archive behind it
3 layers
of checks — metadata, manipulation, reality
20 yrs
of hourly weather behind the cross-check
Denser coverage
our own stations on top of official networks
40+
weather models in the blend
Pairs with the Claim Assessment Tool — one validates the event, the other the evidence
More products

One engine, many tools

Bring us your trickiest closed case.

Send a handful of settled claims — including the ones that turned out to be fraud — and we’ll show you what the screening would have flagged, and why.

See a screening demo