kai-topical-map

name kai-topical-map description Build an AEO-first topical map optimized for AI search citation — entity clusters, query fan-out coverage, information gain scoring, and multi-platform distribution. Produces entity map, content node architecture, schema blueprint, and 90-day publishing calendar. Use when "topical map", "content architecture", "site structure", "topic clusters", "pillar content plan", "AEO map", "AI search architecture", "entity map", "what content should we build", or any request to plan a site's topical structure for AI search visibility. Build a topical map optimized for AI search engines and traditional search together. Every decision is driven by retrievability, entity clarity, source quality, and measurable user demand, not by promises that a specific AI engine will cite a page. Why This Matters AI search visibility is sampled, volatile, and engine-specific. Treat any traffic, citation, or conversion benchmark as context until it is verified for the client with a source, retrieval date, evidence tier, and confidence label. This skill builds the map that can be measured: pages to create, entities to clarify, passages to make retrievable, sources to cite, and follow-up checks to run. Phase 0: Load Product Context Check if MARKETING.md exists in the project root (same directory as CLAUDE.md, README.md, package.json). If it exists: Read it — skip product discovery questions. It has the product name, ICP, value prop, monetization, brand voice, current channels, and competitive landscape. If it does NOT exist: Auto-explore the codebase to create it in the project root (next to CLAUDE.md). Do NOT ask the user what the product is. Read CLAUDE.md, README.md, PROJECT.md, package.json, landing pages, and any project files. Search for email/ad/analytics config. Then create MARKETING.md using the template from /kai-email-system . Present draft to user for confirmation. Phase 1: Topic Space Discovery Load before starting: knowledge/frameworks/aeo-ai-search/aeo-ai-search-playbook-2026.md Read from MARKETING.md . Only ask about things not covered there: Brand/product — what entity should AI engines associate with this topic space? Topic space — the broad domain (e.g., "AI receptionist software", "personal injury law", "B2B cold email") Existing content inventory — blog post URLs, podcast episode titles, landing pages, guides already published Competitor URLs — 2-3 sites currently winning AI citations in this topic space Target queries — what questions should AI answer with YOUR brand? (e.g., "What's the best AI phone answering service?") Current AI presence — sample 3-5 category queries in ChatGPT, Perplexity, Bing/Copilot, and Google AI surfaces where available. Record prompt, location, date, engine, account state, citations, mentions, and missing-data caveats. Output: workspace/topical-map/_discovery.md Baseline AI Presence Scorecard Query ChatGPT Perplexity Google AI Brand Mentioned? Who IS Mentioned? [category query 1] Y/N [competitors] [category query 2] Y/N [competitors] ... This scorecard becomes the "before" measurement. Run it again at 30/60/90 days to track progress. Phase 2: Entity Map Load before starting: knowledge/frameworks/aeo-ai-search/entity-seo-knowledge-graph-deep-dive.md Map the Entity Landscape Identify every entity the brand needs to establish, strengthen, or associate with in the Knowledge Graph. Organize into three tiers: Tier 1 — Own: Entities the brand must be the canonical authority for. These get Entity Home pages. Brand name, product name, founder/CEO, proprietary methodology or framework names Each Tier 1 entity needs: Entity Home URL, Wikidata QID (or submission plan), Schema.org markup Tier 2 — Associate: Entities the brand should appear in context with. These become content cluster topics. Industry terms, use cases, methodologies, problem categories, competitor categories Appearing alongside these entities builds "context vectors" (Bill Slawski's patent analysis) that signal expertise Tier 3 — Reference: Authoritative entities to cite for credibility amplification. Research institutions, industry standards, regulatory bodies, recognized experts Cite these entities to improve provenance and passage usefulness. Do not promise a fixed citation or visibility lift from citations. Entity Map Output | Entity | Type | Current KG Status | Tier | Entity Home URL | Wikidata QID | Schema Types | Action | |--------|------|-------------------|------|-----------------|--------------|--------------|--------| Schema.org Blueprint Generate a @graph structure for the site showing: Organization or Person as the root entity with sameAs links (Wikidata, LinkedIn, Crunchbase, social profiles) knowsAbout properties listing Tier 2 entities mentions and about properties connecting content pages to entities sameAs links for every entity that has a Wikidata or Wikipedia entry Evidence requirement: Treat Knowledge Graph, Wikidata, and vector-search claims as source-dependent. Add source URL, retrieval date, evidence tier, and confidence before using any quantitative claim in a client-facing map. Gate: Every Tier 1 entity must have a proposed Entity Home URL and a Wikidata action plan. Output: workspace/topical-map/_entity-map.md Phase 3: Query Fan-Out Coverage Matrix Load before starting: knowledge/frameworks/aeo-ai-search/query-fan-out-guide.md knowledge/frameworks/content-copywriting/qdp-qdh-qds-content-architecture.md How AI Search Decomposes Queries Google's AI Mode uses "Query Fan-Out" (Liz Reid, Google I/O 2025): complex queries are broken into ~8 simultaneous sub-queries, results are retrieved in parallel, then synthesized into a single answer with layered citations. To rank for a pillar topic, your content must satisfy multiple sub-intents. Build the Matrix For each pillar topic identified from Phase 2's entity clusters: Identify the primary query — the question a user asks AI about this topic Decompose into ~8 sub-queries using PAA mining (3-4 levels deep): Mine "People Also Ask" for the primary query Click through PAA 3-4 levels to discover recursive sub-topics Categorize each sub-query by intent facet: Facet Example for "AI phone answering" Definition "What is an AI phone answering service?" Cost "How much does AI phone answering cost?" Process "How does AI phone answering work?" Comparison "AI phone answering vs live receptionist" Safety/Risk "Are AI phone answering services reliable?" Timeline "How long to set up AI phone answering?" Alternatives "Best AI phone answering services 2026" Technical "AI phone answering integrations with CRM" Map existing content against sub-queries: | Pillar | Primary Query | Definition | Cost | Process | Comparison | Safety | Timeline | Alternatives | Technical | Coverage | |--------|--------------|------------|------|---------|------------|--------|----------|-------------|-----------|----------| | [topic] | [query] | [covered/partial/gap] | ... | ... | ... | ... | ... | ... | ... | X/8 | Apply QDP/QDH/QDS to each sub-query: QDP (Query Deserves Page) : High demand + distinct intent → dedicated URL QDH (Query Deserves Heading) : Moderate demand → section within hub or spoke page QDS (Query Deserves Sentence) : Low demand → inline mention Gate: Every pillar must have ≥6/8 sub-queries identified. Pillars with <4/8 currently covered = priority gaps. Output: workspace/topical-map/_fan-out-matrix.md Phase 4: Information Gain Audit Load before starting: knowledge/frameworks/aeo-ai-search/patent-information-gain-US12013887B2.md knowledge/frameworks/aeo-ai-search/hidden-aeo-edges.md Why Information Gain Matters Google's Information Gain patent (US12013887B2) calculates content novelty using word2vec embeddings. Content that paraphrases what already exists — even in different words — scores LOW. Content must be "orthogonal to consensus" to earn citations. The math: IG Score = f(V_new, V_history) where high KL-divergence from existing content = high score. Audit Process For each pillar topic: Analyze top 5-10 ranking results — list the facts, angles, data points, and framing that ALL of them share. This is the "consensus content." Identify Information Gain opportunities in five categories: IG Category Description Citation Multiplier Proprietary Data Original research, internal metrics, case studies you own 3.2x more citations (Perplexity study) Contrarian Position Evidence-backed views contradicting consensus Triggers Perplexity entropy diversity signal Experience Gap First-person specifics AI cannot fabricate Required for E-E-A-T "Experience" (QRG 4.6.6) Novel Framing Unique terminology, frameworks, mental models Creates semantic distance from competitors Second-Click Content Optimized for the follow-up query after user views #1 result Captures recursive fan-out queries Score each opportunity: High : Brand has proprietary data or unique experience ready to publish Medium : Angle exists but requires research/data collection first Low : Theoretical advantage but no current evidence to support it Information Gain Map | Pillar | Consensus Content (what everyone says) | IG Angle | IG Category | Novelty Score | Source/Evidence Available | |--------|----------------------------------------|----------|-------------|---------------|--------------------------| Gate: Every pillar must have ≥2 "High" novelty opportunities. Pillars with no proprietary data or unique experience = flag for research/data collection before content creation. Output: workspace/topical-map/_information-gain-audit.md Phase 5: Content Node Architecture Load before starting: knowledge/frameworks/aeo-ai-search/geo-academic-research-synthesis.md knowledge/frameworks/aeo-ai-search/perplexity-ranking-reverse-engineered.md Synthesize Into Hub-and-Spoke Architecture Combine Phases 2-4 into the actual pages to create: Each entity cluster (Phase 2) becomes a Hub Each QDP sub-query (Phase 3) becomes a Spoke Each QDH sub-query becomes a section within a Hub or Spoke Each QDS item becomes a sentence within the relevant page Content Node Specification For each node, define: Identity: URL (clean, entity-descriptive) Title Node type: Hub / Spoke / Entity Home AEO Signals: Primary entity served (from Phase 2) Fan-out queries answered (from Phase 3) Information Gain angle (from Phase 4) Citation Signal Requirements (evidence-safe operating targets): Statistics: at least 3 sourced data points per page when the topic benefits from data Expert quotes: at least 1 permissioned quote or attributed expert source when claims need authority External citations: at least 5 primary or high-quality secondary sources for research-heavy pages Atomic fact density: 2-3 verifiable facts per paragraph Paragraph length: 60-100 words when it improves scannability and passage retrieval Sentence length: 15-20 words maximum Answer position: direct answer in first 30-50 words after H2 Schema Prescription (eligibility and clarity, not guaranteed citation lift): Content Node Type Primary Schema Why Use It FAQ / Q&A content FAQPage where eligible Clarifies question-answer structure How-to / Process guides HowTo where eligible Clarifies steps, tools, and prerequisites Data / Research / Stats pages Dataset where eligible Clarifies dataset ownership and fields All informational content Article or BlogPosting Clarifies authorship, dates, and subject Entity Home pages Organization or Person + sameAs Clarifies canonical entity identity Rule: All schema must have EVERY relevant attribute populated. Generic/incomplete schema produces an 18% citation penalty vs having no schema at all (Growth Marshal, Feb 2026). Citation Impact Score Score each node to prioritize publishing order: Citation Impact Score (1-10) = (Fan-Out Coverage × 0.3) # How many sub-queries does this page answer? + (Information Gain Novelty × 0.3) # How novel is this content vs consensus? + (Entity Authority × 0.2) # Does this page strengthen the entity graph? + (GEO Signal Density × 0.2) # How citation-dense is this page? Where: Fan-Out Coverage: (sub-queries answered / total sub-queries for pillar) × 10