Developer API

Know what you can trust.

One API call returns a normalized 0–1 trust score for any text, image, or video. Ensemble AI detection plus cryptographic C2PA provenance — reconciled into a single answer.

No credit card required. 500 text + 200 image calls free every month.

Request
POST /scan/image/url
X-API-Key: vgk_••••••••

{
  "url": "https://example.com/photo.jpg"
}
Response
{
  "content_type": "image",
  "ai_score": 0.38,
  "deepfake": 0.38,
  "c2pa":     null,
  "report": {
    "verdict":    "mixed",
    "confidence": "low"
  }
}

Content authenticity is broken. The tooling is worse.

AI-generated text, synthetic images, and manipulated video are indistinguishable at scale. Teams end up building and maintaining a patchwork of integrations that drift out of sync, return incompatible scores, and leave entire content types unaddressed. Verigin closes all the gaps with one integration.

Everything you need to trust your content pipeline.

📝

Three content types, one API

Text, images, and video share the same endpoint structure and response schema. Build once, apply everywhere.

🔐

C2PA cryptographic provenance

Checks for signed content credentials on every call. A valid C2PA manifest is the strongest trust signal available — cryptographic, not probabilistic.

🎯

Ensemble accuracy

Multiple detection methods run in parallel internally. Disagreements are reconciled before the response is returned. One confident score, not raw probabilities to interpret.

🔍

Verbose signal breakdown

Enable verbose mode for a full audit trail: which signals fired, how strongly, whether provenance was present, and how the final score was composed.

📊

Consistent 0–1 scoring

The same scale across every content type. Your moderation rules, alert thresholds, and dashboards stay simple regardless of what you're checking.

Production-ready in 10 minutes

One API key, clear docs, and a Sandbox tier with no credit card. Average response time under 800ms for text and image. SDKs for Python and Node coming soon.

One score. Multiple signals. Full transparency.

Every Verigin call runs through two independent verification layers.

🤖 Ensemble AI detection

Multiple detection methods analyze content in parallel. Each has different strengths — some excel at fully synthetic content, others at subtle edits and human-AI hybrids. Verigin reconciles the signals internally and surfaces one confident output. You don't pick a model or manage thresholds. The reconciliation is done for you.

🔐 C2PA cryptographic provenance

C2PA (Coalition for Content Provenance and Authenticity) is an open standard backed by Adobe, Google, Microsoft, and the BBC. When content carries a C2PA manifest — a signed, tamper-evident record of its origin and edit history — Verigin reads and validates it. A valid credential is the strongest possible trust signal because it's cryptographic, not probabilistic.

📊 The trust score

The two layers combine into a single ai_score between 0 and 1. Higher = more likely AI-generated or manipulated. Lower = more likely human origin. C2PA credentials, when present, are returned separately as a provenance object.

EU AI Act Compliance

August 2026: the enforcement deadline is real.

EU AI Act Articles 50 and 52 require organizations to label AI-generated content and maintain transparency for AI systems interacting with the public. Penalties reach €15 million or 3% of global annual turnover. Verigin's API is built for exactly this workflow.

📋

Article 50 — AI content labeling

Organizations deploying AI-generated text, audio, or video in public-facing systems must label it. Verigin's detection API provides the labeling infrastructure. Every call returns an auditable record suitable for compliance documentation.

🔍

Article 52 — Transparency obligations

AI systems interacting with users must disclose their AI nature. Verigin's verbose output mode is designed as a compliance artifact — a full audit trail of which signals fired and how the final score was composed.

📅

Enforcement: August 2026

Organizations in scope cannot defer. Media companies, HR platforms, and social networks operating in the EU need detection and labeling infrastructure now. Enterprise contracts include an EU AI Act compliance guarantee clause.

Organizations in scope
Media companies with EU distribution HR platforms operating in the EU Social networks with EU users Content management platforms in regulated industries
Talk to us about EU AI Act compliance →

Built for the teams that can't afford to get this wrong.

Publishing & editorial

Flag AI-drafted submissions before they reach the editorial queue. Surface C2PA credentials from verified photojournalists automatically.

Trust & safety

Score user-generated content at ingest. Route low-trust content to human review with a full signal breakdown to justify the decision.

Ad verification

Detect AI-generated creative assets before they run. Protect brand safety without slowing down campaign operations.

Compliance & risk

Maintain an auditable record of every content check. Verbose mode output is designed for storage as a compliance artifact.

Disinformation research

Analyze content at scale with consistent scoring across text, image, and video — cross-modal analysis that fragmented tools can't support.

Content moderation

Integrate a trust score into your moderation pipeline. Automate the easy cases and surface hard ones to human reviewers.

Start free. Scale without surprises.

No credit card required to get started.

Plan Sandbox Builder Scale Enterprise
Price Free $49 / mo $199 / mo From $1,500 / mo
Text calls 500 / mo 10,000 / mo 60,000 / mo Custom
Image calls 200 / mo 5,000 / mo 25,000 / mo Custom
Video 120 min / mo Custom
Overage — text $0.006 / call $0.004 / call Negotiated
Overage — image $0.008 / call $0.005 / call Negotiated
Verbose breakdown
C2PA provenance
Uptime SLA 99.5% 99.9% 99.9%
Support Docs only Email (48h) Email + Slack (4h) Dedicated
DPA included
Get free Sandbox key Talk to us about Enterprise →

Built for developers. Not committees.

Clear docs, real examples, and a response schema that doesn't change under you.

Text detection — curl
curl -X POST https://api.verigin.com/scan/text \
  -H "X-API-Key: vgk_••••••••" \
  -H "Content-Type: application/json" \
  -d '{"text": "Your article text here..."}'
Image URL detection — curl
curl -X POST https://api.verigin.com/scan/image/url \
  -H "X-API-Key: vgk_••••••••" \
  -H "Content-Type: application/json" \
  -d '{"url": "https://example.com/image.jpg"}'

Common questions.

Verigin doesn't rely on a single detection method. Multiple independent signals are run and reconciled, which smooths out the noise of any one approach. Verbose mode shows you the breakdown so you can apply your own confidence threshold and tune your pipeline accordingly.
True today. But C2PA is being embedded natively in Adobe tools, Android cameras, major stock photo platforms, and broadcast infrastructure. Verigin checks for it on every call — so as adoption grows, your system benefits automatically without any code changes.
Verigin returns a trust signal, not a verdict. The score is designed to inform human review or downstream logic, not replace it. Verbose mode gives you the evidence behind the score so your team or your audit trail can justify every decision independently.
Verigin does not retain submitted content after scoring. Text is processed in-memory. Images and video are scored and discarded. Enterprise plans include a DPA.
Most in-house approaches cover one content type and miss image and video. Verigin adds C2PA provenance — a cryptographic layer no pure-ML approach can replicate — and extends the same trust signal across all three content types without additional engineering.
We publish false positive rates by content type and update them quarterly. Current rates: ~6% for standard English text, ~9% for AI-generated image detection. Non-native English writing has a higher FP rate (11.4%) and we have an active bias audit underway. See our full methodology page for current numbers, confidence intervals, and the appeals process.

You can be scoring content in 10 minutes.

Get your free Sandbox key and see a trust score come back. No credit card, no meeting required.

Get your free Sandbox key → Try the live demo →