Answer Engine Optimization (AEO / GEO) Playbook

May 29, 2026 · View on GitHub

An open field guide to getting cited by name in ChatGPT, Perplexity, Gemini, Claude, Grok, and Google AI for local and specialty businesses.

Maintained by KailxLabs, the studio that builds AI Citation Foundations for premium specialty businesses. This repository is the open, vendor neutral version of the methodology we use on client engagements. Use it, fork the checklists, ship the schema patterns. If you would rather have it built for you with a citation guarantee, that is what we do at kailxlabs.co.

Why this exists

Buyers no longer scroll ten blue links. They ask an assistant "who is the best fertility clinic near me" or "which personal injury firm should I call after a truck accident" and they act on the names the model returns. If your business is not in that answer, the click never happens and you never know it existed.

Traditional SEO optimizes for a ranking. Answer Engine Optimization optimizes for being the entity the model trusts, retrieves, and quotes by name. Different game, different evidence layer.

This playbook covers what that evidence layer actually is and how to build it.

Core concepts
What counts as a citation
The nine layers of an AI Citation Foundation
Schema.org entity graph patterns
Crawler and manifest files
Per vertical checklists
How to measure AI visibility
Glossary
About KailxLabs

Core concepts

Term	What it means
GEO	Generative Engine Optimization. Making your content the source a generative model pulls from when it composes an answer.
AEO	Answer Engine Optimization. Structuring information so an answer engine can extract a clean, attributable answer about your business.
Answer Capsule	A short, self contained, factual block that answers one buyer question completely, so a model can lift it without stitching context from elsewhere.
Island Test	Read one section of a page in isolation. If it cannot stand alone as a complete answer, a model cannot quote it cleanly.
Share of Model	How often your entity appears across a fixed set of buyer prompts on a fixed set of engines. The AI era version of share of voice.
Fact Density	Verifiable, specific facts per hundred words. Models prefer high fact density sources because they are cheaper to verify.

Full definitions live in the Glossary and on the live KailxLabs glossary.

What counts as a citation

Not every mention is a win. We score three tiers:

Named citation. The model states your business name in the answer body. This is the only tier that reliably produces booked calls.
Linked source. Your domain appears in the sources or footnotes panel (Perplexity, Google AI Mode, Gemini) without the name in the prose. Useful, weaker.
Adjacent mention. Your category or city is described accurately but no specific provider is named. This is the gap state most businesses sit in.

The target is a named citation on a buyer intent query across at least two engines. Read the full definition at what counts as a citation.

The nine layers of an AI Citation Foundation

A citable business is built in layers. Skipping a layer leaves a hole a model falls through.

AI Visibility Audit. Run twenty buyer intent prompts across the major engines and record where the business is named, linked, or absent. This is the baseline.
AI readable website. Service pages, city pages, FAQ sections, comparison tables, proof sections, and booking CTAs written so each section passes the Island Test.
Service specific citation pages. One page per service, each opening with an Answer Capsule that fully answers the buyer question.
Schema.org entity graph. A connected @graph of Organization, LocalBusiness, Service, Person, FAQPage, WebPage, and BreadcrumbList, plus the vertical specific type.
Crawler and manifest setup. llms.txt, agents.md, ai.txt, robots.txt, sitemap.xml, and IndexNow configured so AI crawlers are welcomed and guided.
Entity synchronization. Consistent name, address, phone, and description across Google Business Profile, Bing Places, Apple Business Connect, niche directories, and LinkedIn, so the model resolves one entity rather than three conflicting ones.
Citation activation. Directory descriptions, community answers, a long form article, and a review language system that seeds the corroborating sources models cross check.
Lead capture. A booking flow that turns the new high intent traffic into recorded contacts.
AI visibility tracking. Weekly checks against the agreed query set for forty five days, ending in a citation report.

Schema.org entity graph patterns

A loose pile of JSON LD blocks does not help. The win is a connected graph where every node references the others by @id. Minimal skeleton:

{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "Organization",
      "@id": "https://example.com/#organization",
      "name": "Example Specialty Clinic",
      "url": "https://example.com",
      "sameAs": [
        "https://www.linkedin.com/company/example",
        "https://g.page/example"
      ]
    },
    {
      "@type": "MedicalClinic",
      "@id": "https://example.com/#localbusiness",
      "parentOrganization": { "@id": "https://example.com/#organization" },
      "address": { "@type": "PostalAddress", "addressLocality": "Austin", "addressRegion": "TX" }
    },
    {
      "@type": "Service",
      "name": "GLP-1 Weight Management",
      "provider": { "@id": "https://example.com/#localbusiness" }
    },
    {
      "@type": "FAQPage",
      "mainEntity": [
        { "@type": "Question", "name": "How much does the program cost?",
          "acceptedAnswer": { "@type": "Answer", "text": "Programs start at ..." } }
      ]
    }
  ]
}

Vertical specific top types we use: MedicalClinic, Dentist, LegalService, Attorney, HomeAndConstructionBusiness. See the live Schema.org @graph entry for the full pattern.

Crawler and manifest files

File	Purpose
`robots.txt`	Explicitly allow GPTBot, ClaudeBot, PerplexityBot, Google-Extended, and the other AI crawlers. Blocking them by accident is the most common silent failure.
`llms.txt`	A plain text map of your most important pages and facts, written for language models.
`ai.txt`	Declares how AI systems may use your content.
`agents.md`	Instructions for agentic crawlers navigating the site.
`sitemap.xml` plus IndexNow	Standard discovery, plus instant push notification of new and changed URLs.

Reference the live specs: llms.txt, ai.txt, GPTBot, ClaudeBot, PerplexityBot, Google-Extended.

Per vertical checklists

The playbook patterns shift by industry. KailxLabs publishes worked guides for each:

Quick checklist guides: clinics, law firms, home services.

How to measure AI visibility

You cannot improve what you do not measure, and AI answers are non deterministic, so single checks lie. Our method:

Lock a query set. Ten to twenty real buyer prompts, written the way a prospect actually asks.
Lock an engine set. ChatGPT, Perplexity, Gemini, Claude, Grok, Google AI Mode, Google AI Overviews.
Sample, do not snapshot. Run each query several times per engine to account for variance.
Score every run as named citation, linked source, or absent.
Track the trend weekly across forty five days, not a one time screenshot.

Full method: how we measure AI visibility.

Glossary

A condensed version of the full KailxLabs glossary:

Generative Engine Optimization (GEO) — optimizing to be the source a generative model composes its answer from.
Answer Engine Optimization (AEO) — structuring information for clean, attributable extraction.
Answer Capsule — a self contained factual block that fully answers one question.
Island Test — can a section stand alone as a complete answer.
Unresolved Reference Rate (URR) — how often a model references your category but cannot resolve a specific provider.
Schema.org @graph — a connected set of structured data nodes linked by @id.
Retrieval Augmented Generation (RAG) — the retrieve then generate pattern most answer engines use.
Semantic Chunking — splitting content into meaning complete units a retriever can rank.
Share of Model — your entity's presence across a fixed prompt and engine set.
Citation Frequency — how often you are named across repeated runs.
Fact Density — verifiable facts per hundred words.

About KailxLabs

KailxLabs builds the AI Citation Foundation for premium local and specialty businesses. Not a prettier website. The full evidence layer that AI engines use to understand, verify, cite, and recommend a business by name.

What you get: the nine layers above, built end to end.
Price: $5,999 one time. By application.
Delivery: built in ten working days, tracked for forty five.
Guarantee: cited by name on at least one agreed buyer intent query across at least two agreed AI or search engines by day forty five, or the full build fee is refunded. The client keeps the website, code, schema, and content either way.
Engines tracked: ChatGPT, Perplexity, Gemini, Claude, Grok, Google AI Mode, Google AI Overviews.

Answer Engine Optimization (AEO / GEO) Playbook

Why this exists

Table of contents

Core concepts

What counts as a citation

The nine layers of an AI Citation Foundation

Schema.org entity graph patterns

Crawler and manifest files

Per vertical checklists

How to measure AI visibility

Glossary

About KailxLabs

Verticals served

Selected work

Links

License