EU AI Act Article 50 is now in force — AI-generated content must be identifiable. What it means for claims teams →
In commercial pilot with Danish insurers

Seeing is no longer believing.

VeriFly detects AI-generated and AI-edited images and documents with forensic precision — calibrated verdicts in seconds, court-ready reports, and every byte processed inside the EU.

🇪🇺 EU-only processingGDPR · DPA includedC2PA-awareCourt-ready PDF reports
100%
EU processing — no exceptions
6
forensic layers per image
90+
generators in the training corpus
Seconds
from upload to verdict

Trained against the modern generator landscape

FLUXStable DiffusionMidjourneyDALL·EGPT ImageGemini / ImagenFireflyGrokQwen-ImageIdeogramPlaygroundHunyuanDiTSANARecraftLeonardoLumaReveSeedreamFLUXStable DiffusionMidjourneyDALL·EGPT ImageGemini / ImagenFireflyGrokQwen-ImageIdeogramPlaygroundHunyuanDiTSANARecraftLeonardoLumaReveSeedream

Synthetic media has broken the default assumption behind every image-based decision.

~60 seconds

is all it takes to fabricate convincing photo or document evidence with free, phone-native tools.

0 visible traces

remain in modern AI edits — the classic tells are gone, and human review no longer catches them.

Every unverified image

that supports a payout, a verdict or a story is now an unpriced liability.

Technology

Six forensic layers. One calibrated verdict.

No single signal is trustworthy on its own — so VeriFly never relies on one. Independent analysis layers each examine the image from a different angle, and a calibrated fusion step weighs them into a single, explainable verdict.

Pixel committee

An ensemble of specialised vision models reads the image signal itself — trained against fully generated images, AI edits and generative fill inside genuine photos, across 90+ modern generators.

Metadata & provenance

EXIF, camera profiles, GPS, C2PA content credentials, filename patterns and compression traces — including platform-trace recovery when social media has stripped the metadata.

Degradation robustness

Trained on the same material across many levels of compression and resizing, so verdicts hold on images as they arrive from WhatsApp, Messenger and claim portals — where most detectors fail.

Court-ready forensic report

PDF
VerdictAI edit detected
Calibrated confidence97.4%
Chain of custodyVerified · SHA-256
Layer breakdown6/6 documented
Plain-language summaryIncluded

Every analysis produces an audit-ready report: verdict, per-layer findings, metadata traces and a plain-language explanation an investigator can attach directly to the case file.

One API call

REST
# request
curl -X POST https://www.veriflied.dk/api/v1/analyze \
-H "Authorization: Bearer vfk_live_…" \
-F "file=@claim_photo.jpg"
# response · seconds later
{
"verdict": "ai_edit",
"confidence": 0.974,
"report_pdf": "https://…"
}

Drop it into claims intake, KYC or content workflows. Per-organisation API keys, IP allowlists and rate limits are managed from the platform.

Industries

Built for teams where images are evidence

LIVE PILOTS

Insurance

Screen every claim photo and receipt at intake; give investigators forensic reports on flagged cases.

Learn more →

Legal & investigations

Authenticate photographic evidence and detect AI manipulation in discovery material.

Learn more →

Media & newsrooms

Verify user-submitted and agency imagery before publication.

Learn more →

Banking & KYC

Catch fabricated documents and manipulated identity imagery in onboarding flows.

Learn more →

Public sector

Validate imagery and documents behind benefits, permits and casework.

Learn more →

E-commerce & marketplaces

Stop AI-fabricated product photos and fraudulent listings at scale.

Learn more →

Compare

Why teams choose VeriFly

CapabilityVeriFlyGeneric AI detectorsManual review
EU-only processing, DPA includedn/a
AI edits & generative fill in genuine photosPartial
Trained on compressed, social-media-degraded imagesn/a
Metadata, C2PA & platform-trace recoveryPartialPartial
Document forensics (receipts, invoices, certificates)Slow
Court-ready PDF report with audit trailWeeks
Verdict latencySecondsSecondsDays–weeks

Benchmark methodology and matched-FPR comparisons available under NDA — evaluated on your own material.

FAQ

Common questions

How does VeriFly detect AI-generated images?

Every image passes through six forensic layers in parallel: metadata and EXIF analysis, C2PA provenance verification, platform-trace recovery, an ensemble of pixel-level detection models, patch-level forensic analysis, and calibrated signal fusion. Generative models leave measurable statistical patterns in the image signal that camera sensors do not produce — the pixel committee is trained to find them across modern generators and editing tools.

Can it detect AI edits inside genuine photos, not just fully generated images?

Yes — this is the core capability. Generative fill and AI editing, where only a region of a real photo is changed, is the dominant real-world manipulation. VeriFly is trained on fully generated images, AI-edited images and local AI insertions inside genuine photos.

Does detection survive WhatsApp, Messenger and social media compression?

Yes. Compression and resizing destroy many detectors because the subtlest traces get erased. VeriFly's models are trained on the same material across many degradation levels — so verdicts hold on images as they actually arrive, not just on pristine files.

Where are images processed?

Exclusively on servers inside the EU — inference, storage, logs and backups. Veriflied is a Danish data processor and provides a DPA, technical compliance documentation and a full audit trail per analysis.

Is any AI detector 100% accurate?

No — and vendors claiming so should be treated with skepticism. VeriFly returns a calibrated probability together with the concrete forensic findings behind it, so an investigator can see what the verdict rests on and make the final call on an informed basis.

How do we integrate it?

Two ways: the web platform for manual review with exportable PDF reports, and a REST API for automatic screening of every incoming image or document — one call, a verdict in seconds. Access is governed per organisation with API keys, IP allowlists and role-based permissions.

See it work on your own cases.

We set up a closed evaluation on your material — real claim images and documents, processed exclusively in the EU. You get a concrete report, not a sales deck.