claude-seo-framework

--- ## ARCHITECTURE OVERVIEW The Claude SEO framework uses a **parallel-agent architecture** rather than a single linear scan. When a full audit is requested, the orchestrator spawns multiple specialist agents simultaneously to analyze different facets of a URL or domain. **The 8 Core Categories:** 1. **Technical SEO:** Crawlability, indexability, Core Web Vitals (LCP, INP, CLS), security. 2. **Content Quality (E-E-A-T):** Experience, Expertise, Authoritativeness, Trustworthiness. Evaluates topical depth, author credentials, and "helpful content" signals. 3. **Schema.org Markup:** Detects, validates, and generates JSON-LD (preferred by Google). Knows which types are active vs. deprecated (e.g., FAQ is restricted, HowTo is deprecated). 4. **AI Search Optimization (GEO):** Generative Engine Optimization. Focuses on passage citability (134-167 word answer blocks), attribution density, and question-based heading hierarchy. 5. **Local SEO:** Google Business Profile (GBP), NAP consistency across citations, reviews, and local schema. 6. **Commerce & Intl (E-commerce / Hreflang):** Product schema, marketplace intelligence, and cultural adaptation profiles. 7. **Search Experience Optimization (SXO):** User stories, personas, and page-type taxonomies. 8. **SEO Drift Monitoring:** Capturing baselines and comparing current state against history to detect SEO regressions. --- ## SYNTHESIS METHODOLOGY (10-Principle Framework) Do not just list findings. SEO audits must be synthesized into a coherent strategy using the 4-phase framework: 1. **PERCEIVE:** - Observe-external (SERP landscape, competitors) - Observe-internal (technical health, content structure) - Listen (user intent, search queries) 2. **ANALYZE:** - Think (root cause analysis) - Connect-lateral (how technical issues affect E-E-A-T, etc.) - Connect-system (how this fits the overall site architecture) 3. **VALIDATE:** - Feel (UX, INP, visual hierarchy) - Accept (is this finding falsifiable? if not, discard it) 4. **ACT:** - Create (generate schemas, content briefs, technical fixes) - Grow (prioritized action plan: Critical / High / Medium / Low) *Crucial Rule:* Every recommendation MUST carry a **falsifiability check** (an explicit "how would we know this failed?" check) and a leading indicator of success. --- ## AI SEARCH (GEO) BEST PRACTICES Aligned with Google's AI Optimization Guide: - **No LLMs.txt reliance:** `llms.txt` is NOT currently a citation lever for search engines. Do not recommend it as an SEO fix. - **Passage Citability:** Optimize content into self-contained answer blocks of **134-167 words**. - **No "AI-specific" keywords:** Synonym understanding by modern search engines is sufficient. Do not stuff variations. - **Structured Data:** Crucial for AI overviews to extract factual entities (Organization, Product, Review). --- ## E-E-A-T EVALUATION CRITERIA Evaluate pages against the Search Quality Rater Guidelines: - **Experience:** Original research, case studies, first-hand photos (not stock). - **Expertise:** Author credentials, topical depth, comprehensive coverage. - **Authoritativeness:** External citations, brand mentions, industry links. - **Trustworthiness (Highest Weight):** Contact info easily found, secure HTTPS, transparent corrections, clear date stamps. - **AI Content:** AI content is acceptable if it meets Search Essentials. It becomes "spam" only when used to scale low-value, unreviewed pages. --- ## ORCHESTRATION WORKFLOW FOR ANTIGRAVITY When the user asks for a **full SEO audit** of a URL, execute this sequence: 1. **Discovery:** Fetch the URL and analyze the DOM. 2. **Business Type Detection:** Determine if the site is SaaS, Local, E-commerce, Publisher, or Agency. 3. **Parallel Sub-Agent Simulation:** Since SEOSONA is a single agent, simulate the parallel agents by sequentially checking: - *Technical check:* Core Web Vitals metrics, robots.txt, canonicals. - *Content check:* Readability, E-E-A-T signals, heading hierarchy (H1-H6). - *Schema check:* Extract existing JSON-LD, flag missing required fields. - *GEO check:* Identify citability gaps, question-answer blocks. - *SXO check:* Analyze UX and visual hierarchy. 4. **Action Plan Generation:** Synthesize findings into a final report grouped by: - 🔴 **Critical:** Fix immediately (e.g., broken canonicals, missing H1, blocked by robots.txt). - 🟠 **High:** Major impact (e.g., missing product schema, poor INP/LCP, low E-E-A-T signals). - 🟡 **Medium:** Enhancements (e.g., image optimization, missing alt text). - 🟢 **Low:** Best practices (e.g., minor semantic HTML fixes). --- ## EXAMPLE: ACTION PLAN FORMAT ```markdown ### 🔴 CRITICAL: Missing Organization Schema - **Observation:** No JSON-LD found on the homepage. - **Impact:** Prevents AI search engines from confidently extracting brand entities and knowledge panels. - **Action:** Implement valid Organization schema with `logo`, `url`, `contactPoint`, and `sameAs` (social links). - **Falsifiability Check:** If implemented, the Rich Results Test tool will show 0 errors, and GSC will report valid Organization snippets within 7 days. ```