Image provenance API
for humans and agents
Verify Content Credentials without building C2PA yourself.
New to C2PA? Start with What is C2PA?, the verification guide, or the architecture walkthrough.
Supporting signers from
Every platform that accepts images needs to answer three questions
Does this image have valid provenance?
C2PA credential verification. Cryptographically signed manifests from creation through editing. Deterministic — not probabilistic.
Do we trust the signer?
Curated signer reputation. Not all C2PA signatures are equal — a Nikon camera and an unknown self-signed cert require different trust decisions.
What when provenance is absent?
Fingerprint recall and fallback risk scoring. Perceptual hashing recovers stripped credentials. ML signals fill the gap — ranked by reliability.
One API. Three capabilities
POST /v1/verify
Upload an image. Get a clear provenance result.
Structured provenance report with a deterministic decision class. Not "AI Generated (94% confidence)" — actionable outcomes your moderation pipeline can consume directly.
- C2PA manifest parsing + signature validation
- Signer trust lookup with reputation data
- EXIF extraction and metadata analysis
- Perceptual fingerprint matching
{
"decision_class": "verified_synthetic",
"provenance": {
"status": "valid",
"signer": "Adobe Firefly",
"signer_type": "ai_generator",
"trust_decision": "trusted",
"trust_reason": "signer_on_trust_list"
},
"reason_codes": [
"c2pa_valid",
"signer_trusted",
"signer_is_ai_generator"
],
"recommended_action": "auto_label_ai",
"human_summary": "Valid Content Credentials from Adobe Firefly. Label as AI-generated."
}Four decision classes. Zero ambiguity
Deterministic outcomes your moderation pipeline can act on. No confidence percentages.
verified_syntheticValid C2PA provenance confirms this image was generated by an AI tool.
verified_camera_originValid provenance from a hardware camera. Cryptographic proof of capture.
unverified_high_riskNo provenance found. Fingerprint match or ML signals suggest synthetic origin.
unverified_low_riskNo provenance found and no strong supporting or contradictory signal. Treat according to your policy.
We don’t say “AI Generated (94%)”
Signals are not equal. C2PA overrides everything. ML is fallback only.
The hierarchy matters
A confidence score tells you nothing about how an image was made. Real images routinely lose EXIF through social platforms, screenshots, and messaging apps — making metadata absence nearly meaningless as a signal.
C2PA credentials are cryptographic proof from the source. Fingerprint recall matches against a known corpus. ML classifiers fill the gap when nothing else is available — and every response labels them as non-definitive.
Simple pricing
Start free. Pay per request when you need more. Talk to us for volume.
Free
Try the API instantly
- 5 verifications / day without an account
- 100 verifications / month with an API key
- Full REST API access
- JSON provenance reports
Pay as you go
No limits, no subscription
- Unlimited verifications
- Pay per request via x402 protocol
- USDC on Base — low fees, instant settlement
- No API key required
- Works with agents and automated workflows
Volume / Enterprise
Reduced pricing at scale
- Volume discounts
- Custom trust policies
- Dedicated support + SLA
- On-premise deployment options
- Custom integrations
From the blog
Technical deep dives on C2PA, content provenance, and building trust infrastructure.
Regulatory Deadline
EU AI Act applies
August 2, 2026
Transparency obligations require platforms to disclose AI-generated content to users. C2PA is becoming the compliance standard. The window to prepare is closing.