Three Layer Architecture — LLM-LD
LLM-LD / Guide / Three Layer Architecture

Three Layer Architecture

A layered approach to AI readability. Each layer builds on the last. Start simple, add depth over time.

Most approaches to "AI SEO" focus on a single tactic — structured data here, a special tag there. LLM-LD takes a different approach: a layered architecture where each layer builds on the last, adding depth and richness to how AI understands your business.

Start with the foundation (Layer 1). Add relationships and entities (Layer 2). Crown it with a complete AI index (Layer 3). Then connect it all to the discovery network with an AI Discovery Page. Each step makes your site more AI-readable than before.

The Core Principle

The layers build on each other. Layer 1 is the foundation — Schema.org markup that works today. Layer 2 adds depth — entities and relationships that require Layer 1. Layer 3 is the capstone — a complete AI index that ties Layers 1 and 2 together. Start with Layer 1, add the others as you're ready. Each layer makes your site more AI-readable than before.

1

Schema.org JSON-LD

The foundation. Works today.

Standard Schema.org markup on your pages. Uses formats that every major crawler already understands. Start here — it's the baseline everything else builds on.

2

Knowledge Graph + Entities

Relationships, not just facts. Requires Layer 1.

Entity files that define your people, products, services, and locations. A knowledge graph that shows how they connect. Gives AI the relationships, not just the facts.

3

llm-index.json

A website for AI. Requires Layers 1 + 2.

Your website is how humans understand your business. Your llm-index.json is how AI understands it. One file, complete context, with pointers to your entities and knowledge graph.

1

Schema.org JSON-LD

The foundation that works today

Layer 1 is your insurance policy. It uses Schema.org — the structured data format that Google, Bing, and every major search engine already supports. AI crawlers built by OpenAI, Anthropic, Perplexity, and others can read Schema.org too. It's the lingua franca of web structure.

What it includes

  • Schema.org JSON-LD markup on every key page
  • Proper typing (LocalBusiness, Organization, Product, Service, etc.)
  • Complete contact information, hours, and location data
  • Structured service and product descriptions

The AI subdomain (optional but powerful)

For maximum signal, create ai.yoursite.com — a subdomain specifically for AI crawlers. This subdomain mirrors your main content but strips out navigation, ads, and visual design, leaving pure structured content. Think of it as the "reader mode" version of your site, built for machines.

✅

The foundation: Layer 1 is where everyone starts. It works on its own, uses existing standards, and makes your site more AI-readable immediately. No waiting for adoption. No dependencies.

Learn more about setting up an AI subdomain →

2

Knowledge Graph + Entities

Structured relationships. Rich context.

Layer 2 goes beyond individual pages. It defines the entities in your business — people, products, services, locations — and the relationships between them. This gives AI systems context that page-level markup can't provide.

What it includes

  • entities.json — Structured definitions of your key entities (team members, products, services, locations)
  • knowledge-graph.json — How entities relate to each other (this person manages this location, this service uses these products)
  • Optional feeds — Real-time data like inventory, availability, pricing

Why relationships matter

When someone asks an AI "Who's the best dentist in Springfield for kids?", the AI needs to know more than just "Acme Dental exists." It needs to know that Dr. Smith specializes in pediatric dentistry, works at the downtown location, and accepts most insurance. Those are relationships — and Layer 2 makes them explicit.

🔗

Requires Layer 1: Your entities and knowledge graph build on top of your Schema.org foundation. They add depth to the structure you've already created.

Learn more about entities and knowledge graphs →

3

llm-index.json

A website for AI

Think of it this way: your website is how humans understand your business. Your llm-index.json is how AI understands your business.

It's a single file at a well-known path (/.well-known/llm-index.json) that gives AI systems everything they need in one request. No crawling. No parsing HTML. No guessing from fragments. Just complete, structured data — optimized for how LLMs actually process information.

Just like a well-designed website makes it easy for a human to understand what you do and take action, a well-structured llm-index.json makes it easy for AI to understand what you do and recommend you to users.

What it includes

  • Summary — A one-liner description and key facts, written for AI context windows
  • Services — What you offer, structured for matching to user queries
  • Actions — What you want AI to recommend (book, buy, contact, etc.)
  • Contact — Complete contact information in a consistent format
  • Hours — Business hours, holidays, availability
  • Pointers — Links to your entities.json and knowledge-graph.json (Layer 2)

Why a well-known path?

The /.well-known/ directory is a web standard for machine-readable files. It's where you find robots.txt, security.txt, and Apple's app association files. By placing llm-index.json here, AI platforms can check this path automatically — just like they check robots.txt today.

🚀

Requires Layers 1 + 2: Layer 3 builds on your Schema.org markup (Layer 1) and points to your entities and knowledge graph (Layer 2). It's the capstone that ties everything together.

Learn more about llm-index.json →

+

AI Discovery Page: The Connector

How the layers get discovered

The three layers define your AI-readable data. But how do crawlers find that data — and how do you join the network? That's where the AI Discovery Page (ADP) comes in.

The ADP isn't a layer — it's the bridge that connects your layers to the discovery network. It's a page on your site at /ai-discovery that:

  • Links to your structured data files (llm-index.json, entities, knowledge graph)
  • Links to the LLM Disco Network at llmdisco.com
  • Tells crawlers "here's where my AI data lives, and here's how to find everyone else"

Requirements to join

To submit your site to the LLM Disco Network, you need at minimum Layer 1 + an ADP. That's it — Schema.org markup on your pages, plus a discovery page that links to the network. You can add Layers 2 and 3 anytime to deepen your implementation.

The network effect

When a crawler hits any site's ADP in the network, it finds a link to llmdisco.com. There, it discovers the master sitemap of every member site. Find one, discover all.

This is how small sites get discovered. You don't need to wait for AI platforms to find you — you just need one site in the network to get crawled, and everyone benefits.

🔗

The minimum: Layer 1 (Schema.org) + ADP gets you into the network. Add Layer 2 for richer context. Add Layer 3 for the complete AI experience. Each layer you add makes your site more valuable to AI systems.

Learn more about the AI Discovery Page →

Why This Architecture Works

Each layer adds value. Each layer builds on the last. You can start simple and add depth over time — or go all-in from day one. Either way, you're ahead of competitors who are doing nothing.

🏗️

Progressive Enhancement

Start with Layer 1 today. Add Layer 2 when you have richer data. Add Layer 3 when you want the complete AI experience. No wasted work.

⏱️

Works Today, Not Someday

Layer 1 uses existing standards. You don't need to wait for anyone to adopt anything. Your site is more AI-readable the moment you implement.

📈

Compounding Value

Layer 2 makes Layer 1 richer. Layer 3 ties everything together. The more you build, the more valuable each piece becomes.

How the Layers Build Up
AI crawler visits your site
├─→ Reads Schema.org markup on pages
└─→ Layer 1: Basic understanding
├─→ Finds entities.json + knowledge-graph.json
└─→ Layer 2: Rich context + relationships
├─→ Checks /.well-known/llm-index.json
└─→ Layer 3: Complete understanding
└─→ Follows AI Discovery Page → llmdisco.com
└─→ Bonus: Discovers entire network
💡

The bottom line: Layer 1 gets you started. Layer 2 adds depth. Layer 3 gives AI a complete picture — like a website built just for them. Add the AI Discovery Page, and you're connected to the network. Your competitors are doing nothing. You're building the future.

Ready to implement?

Start with the Quick Start guide for a complete Level 1 implementation, or dive into the full spec.

Quick Start Guide → Full Specification →