Tier 3: AI Domination: 20 items for AI Overview, ChatGPT, Claude, Gemini, and Perplexity citation

The AI Domination tier covers llms.txt, aeo.json, declarative-fact density, AI crawler allowlists, Speakable schema, and the patterns that get a site cited in ChatGPT, Claude, Gemini, and AI Overviews.

Tier Explanation: Dominate generative AI engines and LLMs by making every page citation-friendly, machine-readable, and entity-strong. As of 2026, AI search demands extractability, original data, verifiable claims, freshness signals, multimodal pairing, and clear provenance so engines like ChatGPT, Perplexity, Claude, Gemini, Copilot, and Grok cite you accurately. All actions execute on website pages, templates, schema, dynamic feeds, and supporting infrastructure. Tiers 1 and 2 must be in place first.

Related Frameworks

This tier implements the following framework documents in the /Framework/ library. Consult them for canonical reference, audit rubrics, and detailed implementation patterns.

• framework-aicitations.md — ChatGPT, Perplexity, Claude, Gemini, Copilot citation mechanics
• framework-knowledgegraph.md — Wikidata, Wikipedia, Knowledge Panel, sameAs network
• framework-entitysalience.md — Per-page entity engineering, NLP salience scoring
• framework-schema.md — Structured data underpinning AI extraction
• framework-multimodalsearch.md — Multimodal queries combining text/image/voice/video
• framework-agenticaisearch.md — Agentic AI search and autonomous task execution
• framework-infogain.md — Original value AI engines weight when citing

A. AI Answer & Extraction (5)

1. AEO — Answer Engine Optimization

• Place complete standalone answer in the first 80–100 words (TL;DR style above the fold)
• Use short paragraphs (3–4 sentences max), bold key claims, numbered lists, and comparison tables
• Add "Key Takeaways" or "Quick Answer" box at top of every high-intent page
• Include author name, last-updated date, and inline citations for every statistic or claim
• Format answers to match question structure: "What is X?" → "X is [direct definition]"
• Write self-contained 40–60 word answer paragraphs that work as extracted snippets
• Add Question and Answer schema where genuinely applicable
• Validation: Test target queries in ChatGPT, Perplexity, and Gemini — page appears in citation list within 30 days

2. GEO — Generative Engine Optimization

• Lead with unique statistics, original research, or comparison data in first 300 words
• Use authoritative, fluent language optimized for LLM summarization (avoid hedging like "might" or "could")
• Add "Expert Perspective" or "From Our Research" subsections with credentialed authors
• Maintain high factual density — every paragraph should contain at least one citable fact
• Use third-person factual statements ("X has been shown to…") that LLMs can directly quote
• Include "Why It Matters" framing that explains stakes and context
• Pair every key claim with a verifiable source link
• Validation: Page cited as source in AI engine answers, original data points get quoted verbatim

3. CAO — Conversational AI Optimization

• Write in natural Q&A format with H2s phrased as actual questions users would ask
• Create dedicated follow-up question sections at the bottom of each topic
• Use conversational H2s ("What is X?", "How do I Y?", "Why does Z happen?")
• Match exact phrasing from voice search queries and chatbot prompts
• Write in second person with direct, helpful tone
• Add "Related questions" section that mirrors PAA structure from Google
• Include conversation starter prompts users could paste into ChatGPT to find your page
• Validation: Page surfaces as source for follow-up questions in conversational AI sessions

4. RCO — RAG Chunk Optimization

• Structure content into self-contained chunks of 200–400 words that make sense in isolation
• Each chunk leads with a topic sentence summarizing the entire section
• Add semantic chunk boundaries via <section> and <article> tags with id attributes for direct linking
• Avoid mid-paragraph references to other sections ("as mentioned above") — chunks must stand alone
• Include enough context per chunk that a retrieval system can return it as a complete answer
• Use descriptive H2/H3 headings that act as chunk titles (RAG systems use heading hierarchy for relevance scoring)
• Test chunk extraction by copying any single section and asking if it makes sense without surrounding context
• Validation: Each section returns standalone meaningful content when extracted, retrieval tools return relevant chunks for target queries

5. PRO — Prompt Response Optimization

• Identify common prompt patterns users send to ChatGPT, Claude, and Perplexity in your niche
• Build dedicated answer pages for each prompt pattern (e.g., "best X for Y under $Z")
• Match content structure to expected AI response format: comparison tables, ranked lists, pros and cons
• Anticipate follow-up prompts and pre-answer them on the same page
• Use prompt-style headings: "Compare X vs Y", "How to choose between A and B", "Best X for [use case]"
• Include "If you're using AI to research this, here's what to ask" section to prime user prompts toward your content
• Test target prompts weekly across major AI engines to verify your content appears in citation
• Validation: Top 20 prompt patterns in your niche return your content as primary citation

B. Entity & Knowledge Graph (4)

6. EEO — Entity Engine Optimization

• Add consistent Organization and Person schema on every page with stable @id URIs
• Build a central entity hub at /entity/ or /about/ with full markup describing your organization
• Link to authoritative external entity sources via sameAs (Wikidata, Crunchbase, LinkedIn, GitHub)
• Use exact entity name consistently across all pages — no variations or abbreviations
• Cross-reference entities across schemas (Organization → founder Person → published Articles)
• Add additionalType to specify subtype precisely (e.g., LocalBusiness → ProfessionalService → WebDesignAgency)
• Maintain entity attributes: founding date, founders, products, services, awards, publications, areas served
• Validation: Google's Knowledge Graph API recognizes entity, schema validates with cross-page links intact

7. KGO — Knowledge Graph Optimization

• Submit complete Organization + sameAs array in JSON-LD on homepage and About page
• Build relationships with related entities via schema parentOrganization, subOrganization, memberOf
• Claim and complete Google Knowledge Panel via Google Search Console verification
• Build out entity attributes: founding date, founders, awards, headquarters, areas served
• Cross-link to Wikidata Q-ID via sameAs so Google connects your entity to its knowledge graph
• Monitor Knowledge Panel for inaccuracies and submit corrections via Google's panel feedback
• Add Person knowledge graph for founders and key personnel with full sameAs network
• Validation: Knowledge Panel appears for brand search, all attributes accurate, Wikidata linked

8. BLF — Brand Language Feed Optimization

• Create public JSON file at /brand-language-feed.json with official brand facts
• Include: brand name, alternate names, values, product specs, positioning, key messaging, founding details
• Add lastUpdated timestamp and version field for AI freshness validation
• Reference feed from robots.txt and llms.txt so AI crawlers discover it
• Update quarterly via CMS export script — automate via webhook on brand asset changes
• Mirror feed contents in human-readable /brand/ page for both AI and user consumption
• Provide downloadable formats: JSON, YAML, plain markdown for different AI consumption styles
• Validation: Feed file returns 200, validates as JSON, AI engines surface accurate brand attributes when queried

9. WIK — Wikipedia & Wikidata Optimization

• Build out Wikidata entry first — lower notability bar than Wikipedia, foundational for AI training data
• Add structured Wikidata properties: instance of, founded by, founding date, headquarters, official website, industry
• Reference Wikidata Q-ID across all Person and Organization schema via sameAs
• For Wikipedia: build neutral, well-sourced article via experienced editor — never self-edit your own entry
• Maintain consistent facts between Wikidata, Wikipedia, your site, and other authoritative sources
• Link Wikidata entry to LinkedIn, Crunchbase, GitHub, and other authority profiles
• Monitor for vandalism or inaccuracies on Wikipedia and Wikidata pages quarterly
• Validation: Q-ID resolves and references your domain, Wikipedia article (if applicable) is stable and accurate, AI engines pull correct attributes

C. Multimodal & Structured Data (3)

10. MMO — Multimodal Optimization

• Pair every key content section with relevant image, video, or diagram
• Add descriptive alt text, captions, and ImageObject / VideoObject schema with full metadata
• Optimize images for Google Lens and Gemini multimodal recognition (clear subject, good lighting, distinct framing)
• Include VideoObject with transcript, thumbnailUrl, uploadDate, duration, embedUrl, contentUrl
• Add transcripts and chapter markers for all videos to make them indexable and quotable
• Use AudioObject schema for podcasts with full transcript embedded in page
• Cross-reference media across schema (Article → image → ImageObject with caption matching content)
• Validation: Images surface in Google Lens results, video appears in YouTube and Google video search, transcripts indexed

11. AFO — AI Function Optimization

• Add structured function descriptions in JSON-LD for tools, calculators, and processes
• Use HowTo schema for procedural content with full step-by-step HowToStep markup
• Include "How to Use This" section with clear inputs, outputs, and expected results
• Document parameters, prerequisites, and edge cases that AI agents need to invoke functions
• Add SoftwareApplication schema for tools with applicationCategory and operatingSystem
• Build agent-friendly descriptions: when to use this tool, when not to, what it returns
• Document API endpoints in OpenAPI spec if you offer one — AI agents may call them directly
• Validation: Tool or function appears in AI assistant responses when users ask procedural questions

12. VEO — Vector Embedding Optimization

• Structure content so semantically related concepts cluster naturally in embedding space
• Use consistent terminology within a topic — don't mix synonyms in the same section
• Define key terms once, then use them consistently throughout the page
• Group related content into clearly bounded sections so vector similarity matches are clean
• Avoid topical drift within a page — each page should map to one tight embedding region
• Use OpenAI or Cohere embedding APIs to test page similarity to target queries before publishing
• Maintain semantic distinctness between competing pages on your site to avoid embedding overlap
• Validation: Cosine similarity between page and target query above 0.7, sister pages clearly distinct in embedding space

D. Crawler Access & Platform (3)

13. ACM — AI Crawler Access Management

• Decide allow/disallow strategy per AI bot — default to ALLOW for citation visibility unless content is sensitive
• Add explicit rules for: GPTBot, ClaudeBot, Claude-Web, PerplexityBot, Google-Extended, OAI-SearchBot, anthropic-ai, ChatGPT-User, Applebot-Extended
• Configure separate rules for training crawlers vs answer crawlers (e.g., disallow GPTBot training, allow OAI-SearchBot for answer surfacing)
• Monitor server logs for AI bot traffic to confirm rules are respected and identify rogue bots
• Block bad-actor scrapers that mimic AI bots via user-agent verification at WAF level
• Add Cloudflare AI Audit or equivalent edge rules for granular bot control
• Maintain bot allowlist documentation for client transparency and audit trail
• Validation: Server logs show expected AI bot traffic patterns, robots.txt directives respected by major bots

14. PSO — Platform-Specific Optimization

• ChatGPT (OpenAI): Direct answers, structured data, clear authority signals; ensure OAI-SearchBot is allowed
• Perplexity: Citation-rich content with diverse, recent sources; ensure PerplexityBot is allowed
• Claude (Anthropic): Nuanced, well-reasoned content with clear caveats and sourcing; ensure ClaudeBot is allowed
• Gemini (Google): Strong E-E-A-T signals, Google ecosystem alignment, claimed Knowledge Panel
• Copilot (Microsoft): Bing visibility prerequisite, IndexNow integration per Tier 1 INO, Bing Webmaster Tools active
• Grok (xAI): Real-time relevance, X/Twitter brand presence, current event tie-ins, recent publication dates
• Test priority queries weekly across all platforms, document where you appear and where you don't
• Maintain platform-specific test query log with citation appearance tracking by date and platform
• Validation: Page appears as cited source in at least 4 of 6 major AI platforms for primary queries

15. CTM — Citation Tracking & Monitoring

• Set up daily citation tracking via tools like Profound, Otterly, AthenaHQ, or Rankability
• Track citation rate per AI engine: ChatGPT, Perplexity, Claude, Gemini, Copilot, Grok
• Monitor share of voice — what percentage of relevant queries cite your site versus competitors
• Track citation accuracy — does the AI quote you correctly or paraphrase you incorrectly?
• Document citation trend lines per page and per topic cluster monthly
• Build feedback loop: pages that aren't getting cited get audited, re-optimized, and re-tested
• Set monthly KPI targets per engine: citation count, share of voice, accuracy rate
• Validation: AI citation dashboard maintained, monthly trend report shows growing citation rate across engines

E. Scale & Dynamic Content (2)

16. PGO — Programmatic Growth Optimization

• Build template-driven page system for scalable answer pages with built-in originality checks
• Auto-generate unique meta titles, descriptions, and schema per programmatic page
• Inject location, product, or attribute variables into deterministic content templates
• Add E-E-A-T elements per template: author byline, last-updated date, citations, trust signals
• Validate every generated page passes thin-content checks before publishing (minimum word count, originality threshold)
• Implement deduplication logic to prevent near-duplicate pages from indexing
• Monitor programmatic page indexation in GSC weekly — high deindexation rate signals templates are too thin
• Validation: 100% of programmatic pages indexed, average time on page above 30 seconds, no thin-content flags

17. DCO — Dynamic Content Optimization

• Use edge or server-side rendering to personalize content based on referrer, geo, or device
• Create modular, parseable content blocks that AI engines can extract independently
• Add <noscript> fallbacks for any JS-rendered content to ensure AI bots see static version
• Use SSR plus hydration model so initial HTML contains full content for crawlers
• Test dynamic variants with Googlebot and GPTBot user-agents to confirm consistency
• Implement A/B testing with proper rel="canonical" handling so AI doesn't see split content as duplicates
• Avoid client-side-only personalization for SEO-critical content blocks
• Validation: View Source shows full content, GSC URL Inspection rendered HTML matches View Source, AI bot user-agent tests succeed

F. Trust & Verification (3)

18. VCO — Verifiable Claims Optimization

• Every factual claim must have a clickable source link or inline citation
• Use Claim and ClaimReview schema for fact-check-eligible content
• Add citation property to Article schema listing primary sources
• Maintain a sources/references section at bottom of every long-form article
• Cite primary sources (research papers, government data, official documentation) over secondary aggregators
• Date-stamp every claim — "as of [date]" for time-sensitive facts
• Add author byline with credentials for every claim-heavy article
• Build trust by providing data files (CSV, JSON) for proprietary research findings
• Validation: Every long-form article has 5+ verifiable citations, ClaimReview validates where applicable, no orphan claims

19. ASM — AI Sentiment Monitoring

• Run brand queries across AI engines monthly: "What is [brand]?", "Is [brand] reliable?", "Best [brand] alternatives"
• Document AI-generated brand descriptions, ratings, and sentiment per platform
• Identify outdated or incorrect AI claims about your brand and prioritize correction
• Counter inaccurate AI claims by publishing corrective authoritative content on owned domain
• Update Wikipedia and Wikidata when AI engines pull from outdated entries
• Track AI sentiment drift over time — note when descriptions change favorably or unfavorably
• Use AI audit tools (Profound, AthenaHQ, Otterly) to automate sentiment tracking
• Validation: Monthly AI sentiment report maintained, no major inaccuracies persist past 30 days

20. FRO — Freshness & Recency Optimization

• Update dateModified in Article schema on every meaningful refresh (not minor typo fixes)
• Add visible "Last updated: [date]" banner at top of every article
• Include current year in titles where relevant ("Best X in 2026")
• Refresh statistics, pricing, and time-sensitive claims quarterly
• Rebuild stale pages over 12 months old that target competitive keywords
• Set up Google Alerts on key topics to surface refresh-worthy news
• Build editorial calendar with refresh slots, not just new content slots
• Add datePublished AND dateModified so AI engines can weight recency appropriately
• Validation: Average article age under 18 months, freshness score (Inlinks or equivalent) above 80, AI engines cite recent content over stale

Summary

• Total items: 20
• Sub-clusters: 6 (AI Answer & Extraction, Entity & Knowledge Graph, Multimodal & Structured Data, Crawler Access & Platform, Scale & Dynamic Content, Trust & Verification)
• Format: Each item includes 6–8 implementation steps plus a validation criterion
• Net change from original: 7 consolidated, 7 added, 6 platform items merged into 1, 1 renamed
• Position in stack: AI tier — depends on Tiers 1 and 2, feeds into Tier 4 (Entity & Authority) and beyond