FOR PUBLISHERS

Reach users through AI agents while maintaining control

AI agents are already browsing your site. ARW ensures they get accurate information, attribute your brand, and can complete transactions—all on your terms.

Read the specification

The publisher control problem

AI companies deploy increasingly sophisticated agents—computer vision, code execution, multimodal browsers—to scrape your content freely. You can't stop them, can't require attribution, and can't monetize their traffic.

What publishers lose today:

✗No control over training vs. inference usage
✗No attribution in AI responses
✗No monetization of agent traffic
✗Content used to train billion-dollar models without compensation
✗Products recommended without brand recognition

How ARW restores publisher control

ARW provides standardized policy declarations plus technical enforcement for actions.

Declare usage policies

Specify training vs. inference permissions, attribution requirements, and rate limits in machine-readable format. While advisory for content, provides legal foundation and accountability.

Require attribution

Declare how your brand should appear in AI responses. ARW headers communicate attribution requirements directly to agents.

Track AI traffic

AI-* headers identify which agents access your content. Monitor behavior, measure impact, and build accountability when policies are violated.

Enforce transactions

Actions require OAuth—agents cannot add to cart, submit tickets, or complete transactions without user authorization. Technically unbypassable.

Enable conversions

Users complete purchases through agents with OAuth consent. Track agent-driven conversions and turn agent traffic into revenue.

Protect business models

OAuth-gated actions ensure transactions flow through your platform. Maintain customer relationships and protect revenue streams.

Real-world publisher scenarios

E-Commerce

Products appear in AI recommendations with attribution. Purchases require OAuth and flow through your platform. Agent traffic becomes a conversion channel.

{
  "training": { "allowed": false },
  "inference": { "allowed": true },
  "attribution": { "required": true },
  "actions": {
    "add_to_cart": { "auth": "oauth2" },
    "checkout": { "auth": "oauth2" }
  }
}

News Publishers

Content helps users via agents, but full articles require website visit. Attribution drives brand awareness and traffic back to your site.

{
  "training": { "allowed": false },
  "inference": {
    "allowed": true,
    "conditions": ["attribution_required", "excerpt_only"]
  },
  "attribution": {
    "required": true,
    "excerptLimit": 150
  }
}

SaaS Platforms

Docs help developers via agents, but trials and tickets require OAuth. Track agent-driven signups for product-led growth.

{
  "training": { "allowed": true },
  "inference": { "allowed": true },
  "attribution": { "required": true },
  "actions": {
    "signup_trial": { "auth": "oauth2", "tracking": true },
    "create_ticket": { "auth": "oauth2" }
  }
}

Progressive implementation

Start with basic policies and add features incrementally. Each step adds value without breaking existing functionality.

Week 1: Add llms.txt with policies

Create basic llms.txt with policy declarations. Define training/inference permissions and attribution requirements.

Weeks 2-4: Create machine views

Build .llm.md files for your top 20 pages. Provide clean, structured content that reduces hallucinations.

Week 5: Add observability headers

Implement AI-* HTTP headers to identify agents and communicate policies in real-time.

Weeks 6-10: Implement OAuth actions

Build action endpoints with OAuth enforcement. Enable transactions through agents with user consent.

Take control of your content in the agent era

Agent traffic is already 10-20% at many sites. By 2026, it could be 30-40%. Establish standards now or accept extraction as inevitable.

Read the specification Back to home