Validated Actor Inventory (VAI)

Protect human yield. Control automation. Create premium supply.

The Future of Web Traffic Demands Transparency

The web is undergoing its most significant shift since mobile. AI-driven consumption is exploding — agents browse, extract, and synthesize content at unprecedented scale. Meanwhile, programmatic buyers are losing confidence in inventory quality as automation floods in alongside human traffic.

The result: Publishers' human impressions are being devalued because buyers can't distinguish them from automation.

This is a transparency concern - not a fraud issue. Today's publishers face:

AI agents that consume content without viewing ads or engaging
Preview tools and renderers that trigger impressions but never convert
Legitimate automation (QA, monitoring, crawlers) that looks "human enough" to slip through
Buyers bidding down globally because they can't tell what's real

Traditional defenses—binary blocking—create a false choice: Either lose legitimate automation (breaking indexing, tools, and access) or let it dilute your human CPMs. Publishers need transparency and control, not just defense.

The web's next chapter requires a new classification system—one that makes traffic origin explicit, machine-readable, and verifiable.

Distinguish Traffic with Confidence-Based Classification

Validated Actor Inventory (VAI) solves this by classifying who or what is generating each impression and how confident the system is in that assessment, at the moment the request is made.

Two Dimensions of Classification

Every impression gets tagged with:

1. Validated Actor Type (VAT) – If content is shown, who will be influenced?

HUMAN - Most likely a real human user
AI_AGENT - AI agents acting on behalf of a human (assistants, research agents, agentic tools)
SHARING - Social sharing and preview bots (link previews, social cards)
OTHER - Non-human automation not intended to influence a person (crawlers, scrapers, monitoring, QA, renderers)

2. Actor Confidence Tier (ACT) – How certain are we?

ACT-1 - High confidence in classification
ACT-2 - Medium confidence
ACT-3 - Indeterminate / low confidence

The VAI Difference

Not binary. VAI doesn't just block or allow—it segments traffic into confidence tiers so publishers can route high-confidence impressions to premium demand, downgrade uncertain traffic to lower-risk buyers, or suppress non-performing automation entirely.

Not opaque. Confidence is explicit. No black-box scores. ACT tiers give you granular control over how each class of traffic participates in monetization.

Not anti-AI. AI agents become a first-class actor type—distinguishable from fraud and from human traffic. This enables future licensing, usage tracking, and premium access models.

Not post-bid. Classification happens server-side, pre-bid, so you can route traffic before it enters the auction.

How This Protects Publisher Revenue

Stop diluting human CPMs
Route high-confidence human traffic (HUMAN + ACT-1) to premium demand paths, protecting your best inventory from being devalued by automation.

Preserve revenue when confidence is low
Suppress ads for known automation, route uncertain traffic to lower-risk demand, and allow legitimate tools and agents to function—preserving indexing, tooling, and access without contaminating premium inventory.

Turn quality into a competitive advantage
Your internal traffic classification becomes a yield-protecting routing layer. Premium buyers get clean supply, and you avoid the false choice between blocking everything or letting automation dilute CPMs.

VAI vs. Pre-Bid IVT Controls

	Pre-Bid IVT Controls	Validated Actor Inventory
Purpose	Exclude invalid traffic	Route traffic to appropriate demand
Publisher posture	Defensive (risk avoidance)	Yield management (protect, route, suppress)
Decision model	Binary (block / allow)	Tiered routing (premium, lower-risk, suppressed)
Confidence	Implicit or hidden	Explicit (ACT-1/2/3)
AI handling	Lumped into "bots"	First-class actor type
Impact on scale	Collapses supply	Preserves scale via routing
Treatment of uncertainty	Often blocked or ignored	Routed to lower-risk demand
Revenue impact	Can reduce via over-blocking	Protects premium while preserving overall yield

VAI complements existing IVT tools—it doesn't replace them. IVT detection identifies risk. VAI classifies and routes traffic. You decide how each confidence tier participates in monetization.

Publisher-Controlled Infrastructure at the Edge

VAI is implemented as publisher-controlled infrastructure that runs at your edge—CDN, gateway, or origin. It provides cryptographically-signed assertions that travel with each request through the programmatic supply chain.

Two Simple Endpoints

GET /pw/vai.json – Returns VAI classification with signed assertion

{
  "iss": "https://paywalls.net",
  "aud": "vai",
  "dom": "example.com",
  "vat": "HUMAN",
  "act": "ACT-2",
  "did": "01J...ULID",
  "iat": 1736629940,
  "exp": 1736630040,
  "kid": "2026-01-a",
  "assertion": "BASE64URL_JWS"
}

GET /pw/vai.js – JavaScript that sets window.__PW_VAI__ for browser integration

Publisher Integration (One Script Tag)

Add to your HTML template:

<script src="/pw/vai.js"></script>

That's it. VAI data is injected at the edge without modifying origin HTML—no cache fragmentation, no client-side complexity.

Header Bidding Integration

VAI data flows into OpenRTB bid requests automatically:

// Header bidding wrapper attaches VAI to bid requests
pbjs.setConfig({
  ortb2: {
    site: {
      ext: {
        pw_vai: {
          iss: "https://paywalls.net",
          aud: "vai",
          dom: "example.com",
          vat: "HUMAN",
          act: "ACT-2",
          kid: "2026-01-a",
          exp: 1736630040,
          assertion: "BASE64URL_JWS"
        }
      }
    }
  }
});

SSP Verification

SSPs verify the cryptographic signature to ensure authenticity:

Assertions are signed with public keys
Bound to the serving domain
Expire quickly (30-120 seconds)
Verifiable via a public JWKS endpoint (prototype path: GET /pw/jwks.json)

Design Principles

Probabilistic, not absolute – Expresses confidence, not certainty
Time-bound – Assertions expire quickly to prevent replay
Verifiable – Cryptographic signatures prove authenticity
Cache-friendly – No HTML variation by actor type
Privacy-preserving – No user identification or tracking
Forward-compatible – Extensible for emerging traffic types
Publisher-controlled – Runs on your infrastructure, bound to your domain

What VAI Enables Next

Because classification is based on actor type and confidence, the same infrastructure supports:

AI agent licensing – Track and monetize AI consumption
Usage-based access – Meter and bill by actor type
Tiered subscriptions – Premium access for verified actors
Wallet integration – Automated payment for agent access

All without rework of the core classification system.

What VAI Is Not

❌ Not a replacement for IVT detection or fraud prevention
❌ Not a promise of fraud-free traffic
❌ Not an identity system for users
❌ Not anti-AI or anti-automation
❌ Not another client-side tag

It is publisher-controlled classification infrastructure for the next phase of the web.

The Bottom Line

If buyers can't distinguish your human traffic, they won't pay for it.

Validated Actor Inventory makes that distinction explicit, verifiable, and actionable—so human attention stops subsidizing automation, and quality becomes your competitive advantage.

Learn More

Try VAI live - See VAI in action with real-time data
Technical spec - Complete implementation details
Get started - Integrate VAI on your site

VAI complements existing IVT detection and fraud prevention systems. It provides an additional layer of transparency for programmatic advertising transactions, not a replacement for existing controls.