THE PROBLEM & SOLUTION

Why ARW?
The Agent-Web Interoperability Gap

AI agents now generate 40% of web traffic (heading to 70% by 2027), but the web wasn't built for them. The result: massive inefficiency, zero observability, and missed opportunities for publishers, agents, and users alike.

The Problem: No Infrastructure

Websites were designed for human browsers, not AI agents. The gaps are costly.

85% Bandwidth Waste

Agents scrape 55KB HTML when 8KB Markdown would suffice. Publishers pay for wasted bandwidth. Agents pay for wasted tokens.

55KB HTML8KB .llm.md

Slow Discovery

Agents must crawl entire sites (5-30 minutes) to understand capabilities. No structured manifest. No guaranteed completeness.

5 min crawl vs 2 sec /llms.txt

Zero Observability

Publishers can't track agent behavior, measure value, or understand impact. Agents are invisible. No analytics. No attribution.

0% visibilityFull AI-* headers

The Cost of No Infrastructure

Publishers

  • • Wasted bandwidth costs
  • • Zero analytics on agent behavior
  • • Missed monetization opportunities
  • • No control over content usage

Agents

  • • High token costs (85% waste)
  • • Slow discovery (minutes not seconds)
  • • Parsing errors and hallucinations
  • • Can't complete transactions

Users

  • • Slow agent responses
  • • Inaccurate information
  • • Incomplete transactions
  • • Poor attribution

The Solution: Infrastructure Layer

ARW provides the missing infrastructure for efficient agent-web interaction. Not adversarial—collaborative. Everyone benefits.

85% Efficiency Gain

  • Machine views (`.llm.md`) reduce tokens by 85%
  • Structured discovery via `/llms.txt` (10x faster)
  • Lower costs for publishers and AI companies

Full Observability

  • `AI-*` headers track agent behavior
  • Foundation for agent analytics platforms
  • Policy compliance monitoring

OAuth Actions

  • Technically enforced transactions
  • Cannot be bypassed by agents
  • Standard OAuth 2.0 flows
MetricWithout ARWWith ARWImprovement
Content Size55KB HTML8KB Markdown85% reduction
Discovery Time5-30 minutes2 seconds10x faster
Observability0% (blind)100% (AI-* headers)Full visibility
ActionsNoneOAuth-protectedTransactional
Bandwidth Cost$450/month$68/month$382/month saved

The Economic Foundation: Coasean Singularity

ARW reduces transaction costs for AI agents by 80-90%, enabling the shift identified in NBER research as the "Coasean Singularity"—the transition to near-zero transaction costs.

Transaction Cost Reduction

Search Costs↓ 90%

Structured discovery via `/llms.txt` replaces crawling

Communication Costs↓ 85%

Machine views vs HTML parsing

Contracting Costs↓ 80%

Declarative OAuth actions

Verification Costs↓ 90%

Observability headers enable tracking

Why This Matters

When transaction costs approach zero, economic organization fundamentally reshapes. ARW provides the infrastructure that enables this transition for AI agents.

Investment Thesis

"ARW is essential infrastructure for the agent economy, not an optional enhancement. As agent traffic grows from 40% to 70%, websites without ARW will be at a severe competitive disadvantage."

PROGRESSIVE LAYERS

Four layers of the new web

"From findable to understandable — and finally operable."

1

Discovery

agents can find your content

2

Semantic

agents can understand your content

3

Actionable

agents can act safely and with consent

4

Interoperable

agents and sites speak shared protocols

The web layer remains irreplaceable as the universal entry point, discovery mechanism, and fallback for the agentic ecosystem — regardless of which specialized protocols emerge.

CORE PRINCIPLES

Ethics encoded in HTTP

Human parity

no SEO penalties

Transparency

visible policies, visible intent

Consent

OAuth before action

Attribution

credit where due

Pragmatism

built on the web's DNA

Ready to Implement ARW?

Start with ARW-1 (Discovery Ready) in just 2 hours

Next: Quick Start

2-hour implementation guide

Compare: vs llms.txt

Migration path and features

Learn: Case Studies

Real-world implementations