AI Discovery Protocol (ADP) Overview
What is the AI Discovery Protocol?#
The AI Discovery Protocol (ADP) is a comprehensive specification that defines how websites can make themselves fully discoverable by AI systems. While individual files like llms.txt and robots.txt each address one aspect of AI visibility, ADP brings them all together into a unified framework with 20 endpoints across 4 compliance levels. ADP was designed to solve a fundamental problem: AI systems need structured, machine-readable information about websites, but there was no standard for how to provide it. Each AI company was building its own crawling infrastructure with different expectations. ADP creates a common language between website owners and AI systems. The protocol covers everything from basic identity (who are you?) to real-time content feeds (what is new?) to proof infrastructure (can we verify your claims?). For most websites, implementing ADP Level 1-2 takes a few hours and provides the majority of the AI visibility benefits.
The Four Compliance Levels#
ADP defines four progressive compliance levels, each building on the previous one. This tiered approach lets you start simple and add sophistication over time.
- Level 1 (Minimal): Add /ai-discovery.json with basic site identity. Takes 15 minutes. Gets you on AI radar.
- Level 2 (Standard): Add /knowledge-graph.json, /llms.txt, and AI-friendly robots.txt. Takes 2-4 hours. Provides substantial AI visibility.
- Level 3 (Advanced): Add versioning, /updates.json feed, and /feed.json for real-time content. Takes 1-2 days. Keeps AI systems up to date.
- Level 4 (Enterprise): Add /news/* namespace and proof infrastructure for verified claims. Takes 1-2 weeks. Maximum trust and authority.
ADP Scoring#
The ADP score is a 0-100 composite metric calculated from three components. The Discovery score (0-50) measures how well AI systems can find your content through llms.txt, robots.txt, and Schema.org markup. The Action score (0-50) measures whether AI agents can interact with your services through MCP, UCP, ACP, and A2A endpoints. The combined score gives a holistic view of your AI readiness. Most websites score between 5-25 on their first scan, with easy improvements available in llms.txt and robots.txt configuration that can boost the score by 20-40 points in an afternoon.
Most websites score 5-25 on their first ADP scan. Adding llms.txt and configuring robots.txt for AI crawlers can boost your score by 20-40 points in a single afternoon.
Getting Started with ADP#
The fastest path to ADP compliance starts with three files: ai-discovery.json for identity, llms.txt for content summary, and robots.txt for crawler access. This gets you to Level 2 compliance, which is sufficient for most businesses. Run a free scan at citability.ai/scan to see your current score and get specific recommendations for improvement.
- 1Run a free scan at citability.ai/scan to see your current ADP score.
- 2Create an llms.txt file describing your site's purpose and key pages.
- 3Update robots.txt to explicitly allow all major AI crawlers.
- 4Add Schema.org JSON-LD markup to your key pages.
- 5Create an ai-discovery.json file with your site's identity information.
- 6Re-scan to verify your improvements and get next-step recommendations.
Frequently Asked Questions
ADP is a specification developed by Citability that is gaining adoption across the web. While it is not yet a W3C standard, it builds on established standards (Schema.org, robots.txt) and is designed to complement emerging standards like MCP and A2A.
Most businesses should target Level 2, which covers the essentials: ai-discovery.json, llms.txt, robots.txt, and Schema.org markup. Level 3-4 are recommended for content-heavy sites, news publishers, and enterprises that need real-time AI feeds.
Level 1 takes about 15 minutes. Level 2 takes 2-4 hours for a developer. Level 3 takes 1-2 days. Level 4 is a 1-2 week project for enterprises. Most of the value comes from Level 1-2.