seo-engineering

# SEO Engineering Engineer search visibility through technical optimization, content strategy, structured data, and generative engine optimization (GEO). This skill covers the full SEO stack from crawlability to AI citation optimization. ## SEO Architecture ``` SEO System ├── Technical SEO (how search engines crawl and index your site) │ ├── Core Web Vitals (LCP, INP, CLS) │ ├── Crawlability (sitemap, robots.txt, internal linking) │ ├── Indexability (canonical URLs, noindex, hreflang) │ └── Site Architecture (URL structure, breadcrumbs, pagination) ├── Content SEO (what you publish and how it ranks) │ ├── Keyword Research (intent mapping, volume, difficulty) │ ├── Content Clusters (pillar pages + supporting articles) │ ├── Programmatic SEO (templated pages at scale) │ └── Content Freshness (update cadence, topical authority) ├── Schema Markup (structured data for rich results) │ ├── Organization, Product, FAQ, HowTo, BreadcrumbList │ ├── Article, Course, SoftwareApplication │ └── Review, AggregateRating, VideoObject └── GEO (Generative Engine Optimization) ├── AI Citation Optimization (ChatGPT, Claude, Perplexity) ├── Entity-first content architecture └── Structured knowledge format ``` ## Core Web Vitals Targets | Metric | Good | Needs Improvement | Poor | |--------|------|-------------------|------| | LCP (Largest Contentful Paint) | < 2.5s | 2.5-4.0s | > 4.0s | | INP (Interaction to Next Paint) | < 200ms | 200-500ms | > 500ms | | CLS (Cumulative Layout Shift) | < 0.1 | 0.1-0.25 | > 0.25 | ### Framework-Specific Optimizations | Framework | LCP Fix | INP Fix | CLS Fix | |-----------|---------|---------|---------| | Next.js | Use `next/image` with priority, SSR above-fold | React Server Components, reduce client JS | Reserve space with `width`/`height` on images | | Vite/React (SPA) | SSG/prerender critical pages, lazy load below-fold | Code split routes, defer non-critical JS | Fixed layout containers, font-display: swap | | Astro | Static by default (good), optimize images | Minimal JS by design (good) | Use Astro Image component | ## URL Structure Best Practices ``` Good: /blog/how-to-learn-solidity /courses/blockchain-fundamentals /tools/smart-contract-auditor /pricing Bad: /blog/post/12345 /p?id=abc123 /courses/course-detail?slug=blockchain /page/about-us/company/info ``` Rules: - Lowercase, hyphens (never underscores or camelCase) - 3-5 words max per slug - Include primary keyword - No dates in evergreen content URLs - Flat hierarchy (max 3 levels deep) ## Sitemap Configuration ```xml <?xml version="1.0" encoding="UTF-8"?> <urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9"> <url> <loc>https://example.com/</loc> <lastmod>2026-04-14</lastmod> <changefreq>weekly</changefreq> <priority>1.0</priority> </url> <url> <loc>https://example.com/blog/learn-solidity</loc> <lastmod>2026-04-10</lastmod> <changefreq>monthly</changefreq> <priority>0.8</priority> </url> </urlset> ``` ### Sitemap Rules - Maximum 50,000 URLs per sitemap file (use sitemap index for more) - Only include canonical URLs (no duplicates, no redirects) - Update `lastmod` only when content actually changes - Exclude noindex pages, paginated pages, search results - Submit via Google Search Console and reference in robots.txt ## robots.txt Template ``` User-agent: * Allow: / Disallow: /api/ Disallow: /dashboard/ Disallow: /settings/ Disallow: /admin/ Disallow: /_next/data/ Disallow: /auth/ Sitemap: https://example.com/sitemap.xml ``` ## Keyword Research Framework ### Intent Classification | Intent | Signal Words | Content Type | Conversion Potential | |--------|-------------|-------------|---------------------| | Informational | how to, what is, guide, tutorial | Blog posts, guides | Low (top of funnel) | | Navigational | [brand name], login, pricing | Homepage, product pages | Medium | | Commercial | best, vs, review, comparison, top 10 | Comparison pages, listicles | High | | Transactional | buy, signup, free trial, discount, pricing | Landing pages, pricing | Highest | ### Content Cluster Model ``` Pillar Page: "Complete Guide to Smart Contract Development" ├── Cluster: "Solidity Tutorial for Beginners" → links to pillar ├── Cluster: "Smart Contract Security Best Practices" → links to pillar ├── Cluster: "Testing Smart Contracts with Hardhat" → links to pillar ├── Cluster: "Smart Contract Design Patterns" → links to pillar └── Cluster: "Deploying Smart Contracts to Mainnet" → links to pillar ``` Each cluster page links back to the pillar. The pillar links to all clusters. This creates topical authority. ## Environment Variables ```bash GOOGLE_SEARCH_CONSOLE_SITE=https://example.com # Verified property URL GOOGLE_SITE_VERIFICATION=XXXXXXXXXXXX # Meta tag verification code ``` ## Measurement ### Key SEO Metrics | Metric | Source | Frequency | |--------|--------|-----------| | Organic Traffic | Google Analytics / PostHog | Weekly | | Keyword Rankings | Google Search Console | Weekly | | Indexed Pages | Google Search Console | Monthly | | Core Web Vitals | PageSpeed Insights API | Monthly | | Backlink Profile | Ahrefs / Moz API | Monthly | | AI Citation Rate | Manual checks + Perplexity | Monthly | ## Citation Science (April 2026) Research from the SEO/GEO Atlas (7 independent passes, April 2026) reveals that AI citation is predictable and engineerable. The following 12-point checklist determines whether a page will be cited by ChatGPT, Perplexity, Google AI Overviews, or Claude. ### The 12-Point Citation Checklist Every page targeting AI citation must pass these checks: 1. **Retrieval Rank** -- Position 0 in retrieval = 58% citation probability. Position 10 = 14%. Rank in the retrieval index is the single strongest predictor of citation. Optimize for traditional search ranking first. 2. **Title-Query Alignment** -- Pages where the title aligns with the query at 50%+ see a 2.2x citation lift. Titles must mirror the exact phrasing users (and AI fan-out queries) use. 3. **Focused Scope** -- 500-2,000 words with 7-20 subheadings. Focused pages beat "ultimate guides." AI systems prefer extractable, self-contained answers over sprawling content. 4. **The Ski Ramp** -- 44.2% of all citations come from the first 30% of a page. Front-load the highest-density, most quotable information at the top. 5. **Optimal Passage Length** -- 134-167 words per section (between H2/H3 headings). 120-180 words is the sweet spot for ChatGPT-style extraction. 6. **Definitive Language** -- Use authoritative, declarative statements. "X is the leading..." not "X might be considered..." 7. **Q&A Format** -- Question headings with direct answers in the first sentence. AI systems extract Q&A pairs with high fidelity. 8. **Entity Density** -- Target 20.6% proper noun density (named people, companies, products, technologies). Web average is 5-8%. High entity density signals factual, citable content. 9. **Balanced Tone** -- Avoid promotional or hyperbolic language. AI systems prefer neutral, authoritative content. A balanced perspective actually increases citation probability. 10. **Simpler Writing** -- Grade level 16 (Flesch-Kincaid). Academic-level clarity without unnecessary complexity. Readable but substantive. 11. **Source Citations** -- Include outbound references to authoritative sources with specific data. AI systems prefer content that itself cites sources. 12. **Freshness Signals** -- Visible datePublished and dateModified. AI systems inject "2026" into 28.1% of fan-out queries, so dated content gets filtered out. ### The Bimodal Reality Only 15% of retrieved pages ever get cited. Citation is bimodal: pages are either always-cited or never-cited. There is almost no middle ground. Passing the 12-point checklist moves a page from the "never" category to "always." ## Platform-Specific Optimization Each AI platform has different source biases, content preferences, and citation patterns. Optimize for the platform your audience uses most. ### ChatGPT - **Source bias**: Wikipedia (48%), Reddit (13%), YouTube (9%) - **Content format**: 120-180 word sections, straight H2 headings, no fancy formatting - **Naming**: Names brands in 36% of responses - **Strategy**: Mirror Wikipedia's neutral encyclopedic tone. Use clear H2 headings that match common queries. Keep sections self-contained and extractable. ### Google AI Overviews - **Source bias**: YouTube (23%), Reddit (21%), established publishers - **Content format**: Schema-dense pages, Google top-10 rank required as prerequisite - **Strategy**: You must already rank in Google's top 10 to be eligible for AI Overviews. Add Article schema with datePublished/dateModified, FAQPage schema, and Organization schema with sameAs links. ### Perplexity - **Source bias**: Reddit (47%), academic/white-paper format content - **Content format**: Data tables, numbered lists, statistical citations - **Strategy**: Write in a white-paper format with explicit data tables. Perplexity heavily favors content that looks like research. Include statistical claims with source attributions. ### Claude - **Source bias**: Favors deep structural content, technical documentation - **Content format**: Detailed, well-organized technical content - **Naming**: Names brands in 46% of responses (vs ChatGPT's 36%) - **Strategy**: Deeper structural content with clear hierarchies. Claude citations favor pages that demonstrate domain expertise through comprehensive coverage. ### The 11% Overlap Problem Only 11% of domains are cited by BOTH ChatGPT and Google AI Overviews. Optimizing for one does not guarantee visibility on the other. Platform-specific pages or content variations may be necessary for full coverage. ## AI Crawlers and robots.txt AI systems use two types of crawlers: **retrieval crawlers** (fetch pages to answer queries in real-time) and **training crawlers** (fetch pages to train models). The recommended strategy is to allow retrieval crawlers (they drive citations) while optionally blocking training crawlers. ### The Balanced robots.txt Configuration ``` # === AI RETRIEVAL CRAWLERS (ALLOW -- these drive citations) === User-agent: OAI-SearchBot Allow: / User-agent: ChatGPT-User Allow: / User-agent: Claude-SearchBot Allow: / User-agent: PerplexityBot Allow: / User-agent: Applebot-Extended Allow: / # === AI TRAINING CRAWLERS (OPTIONAL BLOCK -- your content, your choice) === User-agent: GPTBot Disallow: / User-agent: ClaudeBot Disallow: / User-agent: CCBot Disallow: / User-agent: Google-Extended Disallow: / # === STANDARD SEARCH CRAWLERS (ALLOW) === User-agent: Googlebot Allow: / User-agent: Bingbot Allow: / User-agent: * Allow: / Disallow: /api/ Disallow: /dashboard/ Disallow: /admin/ Disallow: /auth/ Sitemap: https://example.com/sitemap.xml ``` **Key distinction**: `OAI-SearchBot` (retrieval, drives citations) vs `GPTBot` (training). `Claude-SearchBot` (retrieval) vs `ClaudeBot` (training). Blocking training crawlers does NOT affect your citation visibility. ## The LATAM GEO Arbitrage For products targeting Latin America (like DojoOS), there is a 12-24 month arbitrage window in AI-driven search. ### The Global Spanish Problem AI models collapse all Spanish-speaking markets into a single "Global Spanish" bucket. A query about "bootcamps de programacion en Costa Rica" returns the same results as "bootcamps de programacion en Mexico" -- even though the markets, prices, regulations, and audience are completely different. ### The Hreflang Opportunity Proper hreflang tagging (es-CR, es-MX, es-CO, es-AR) is massive arbitrage: - Forces AI systems to recognize market-specific content - Creates differentiated entities per country in the knowledge graph - Extremely low competition (almost nobody in LATAM is doing GEO-specific optimization) - Costa Rica is a Tier-3 market for AI citations -- virtually no competition ### The Window - 12-24 months before larger competitors catch up - First-mover advantage in LATAM GEO compounds (each citation increases future citation probability) - 43% of non-English fan-out queries are actually in English (the "Global English Gatekeeper" effect), so English content targeting LATAM keywords also gets cited ### Implementation 1. Create country-specific landing pages with proper hreflang tags 2. Include country-specific entities (local companies, local regulations, local salary data) 3. Register the brand entity on Wikidata with country-specific claims 4. Submit to Bing Webmaster Tools (Bing powers many AI retrieval indexes) ## Knowledge Base - `skills/seo-engineering/rules/content-strategy.md` -- Keyword research, content clusters, programmatic SEO - `skills/seo-engineering/rules/technical-seo.md` -- Crawlability, indexability, Core Web Vitals - `skills/seo-engineering/rules/geo-optimization.md` -- GEO patterns, AI citation optimization - `knowledge/seo-geo-atlas-2026.md` -- SEO/GEO/AEO/LLM Visibility Atlas April 2026: Citation science, platform architecture, AI crawler configuration, LATAM GEO arbitrage, 90-day action plan. 2,221 lines from 7 independent research passes.