SEO & GEO Architect

Optimize for traditional search AND AI-powered engines. Based on Google Search Central, Moz, and GEO research (2025-2026).

Sub-Commands

Command	Description
audit-seo	Run full SEO audit (titles, meta, schema, speed, mobile)
audit-geo	Run GEO audit (llms.txt, entity clarity, AI answer format)
schema	Generate JSON-LD structured data for any page type
optimize	Implement priority fixes ranked by impact
llms	Generate/update llms.txt for AI crawler discovery

Workflow

Step 1: Run SEO audit

Check: title tags (< 60 chars, keyword first), meta descriptions (< 155 chars, value prop), H1 (one per page, keyword), img alt text, OG tags, canonical, robots.txt, sitemap.xml, Core Web Vitals.

Score: 90-100% good. 70-89% fix flagged. < 70% structural issues first.

Step 2: Run GEO audit

Check: llms.txt presence, ai-plugin.json, FAQ/HowTo schema, entity clarity, brand mention structure, answer format (direct answer in first 2 paragraphs).

Step 3: Prioritized fixes

Fix	Impact	Effort
Fix crawl errors (GSC)	High	Low
Improve LCP < 2.5s	High	Medium
Add structured data (JSON-LD)	High	Medium
Create llms.txt	High	Low
Fix meta titles/descriptions	Medium	Low
Add hreflang (if multilingual)	Medium	Low
Improve internal linking	Medium	Medium

Step 4: Structured data (must implement)

{
  "@context": "https://schema.org",
  "@type": "Organization",
  "name": "Brand Name",
  "description": "One sentence. What you do.",
  "url": "https://example.com",
  "logo": "https://example.com/logo.png",
  "sameAs": ["https://twitter.com/handle", "https://github.com/handle"]
}

Also: Article, FAQPage, HowTo, Product, BreadcrumbList, LocalBusiness (if applicable).

Step 5: llms.txt

# Brand Name
> One-line description.

## Core pages
- https://example.com/: Home
- https://example.com/about: About

## Docs
- https://docs.example.com/api: API reference

## Blog
- https://example.com/blog/post: Title

Place at /.well-known/llms.txt. This is how AI crawlers (GPTBot, Claude, Gemini) discover your site.

Step 6: GEO optimization

Factor	Implementation
Citability	Publish original research, unique data. AI cites sources.
Entity clarity	Clear about pages. Define who/what/for whom.
Answer format	Direct answer in first 2 paragraphs before expanding.
Conversational tone	Write naturally. AI penalizes keyword stuffing.
Source authority	Link to .edu, .gov, peer-reviewed. Build topical authority.
Brand mentions	Get mentioned on authoritative sites.

Production Checklist

• Technical SEO > 90%
• GEO audit passes (llms.txt, schema, entity clarity)
• Core Web Vitals pass (LCP < 2.5s, INP < 200ms, CLS < 0.1)
• Structured data: Organization + page-specific types
• Sitemap.xml submitted to GSC
• robots.txt correct (not blocking anything important)
• hreflang configured (if multilingual)
• Canonical URLs on every page
• Open Graph + Twitter Cards on every page
• llms.txt published
• GSC monitoring active

Anti-Patterns

Mistake	Fix
Keyword stuffing	Natural language. For humans first, search second.
No structured data	Schema critical for Google AND AI search.
No original research	AI cites original data. Publish unique findings.
Thin content (< 500 words)	Comprehensive, authoritative.
Ignoring Core Web Vitals	Performance is a ranking factor.
No E-E-A-T signals	Author bios, citations, credentials, about page.

Technical SEO

Core Web Vitals Thresholds

Metric	Good	Needs Work	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

Measure with: Lighthouse, PageSpeed Insights, Web Vitals extension, CrUX dashboard.

Crawl Budget Optimization

User-agent: *
Allow: /
Disallow: /api/
Disallow: /*?*
Disallow: /search
Sitemap: https://example.com/sitemap.xml

Rules:

• Block faceted navigation URLs with ? parameters
• Block internal search pages (no SEO value, consumes budget)
• Keep sitemap under 50,000 URLs or split into sitemap index
• Set lastmod accurately (Google prioritizes recently-changed URLs)
• Use <priority> sparingly (Google largely ignores it)

Hreflang for Multilingual Sites

<link rel="alternate" hreflang="en" href="https://example.com/en/page" />
<link rel="alternate" hreflang="es" href="https://example.com/es/pagina" />
<link rel="alternate" hreflang="x-default" href="https://example.com/" />

Rules:

• Every language must link to every other language (bidirectional)
• x-default for language selector page or auto-redirect
• Use ISO 639-1 codes (en, es, fr, de, ja, zh)
• Verify with Google Search Console International Targeting report

Schema.org Nesting Patterns

{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "Title",
  "author": { "@type": "Person", "name": "Author" },
  "publisher": { "@type": "Organization", "name": "Site", "logo": { "@type": "ImageObject", "url": "logo.png" } },
  "mainEntityOfPage": { "@type": "WebPage", "@id": "https://example.com/article" }
}

Nest entities, don't flatten. Google prefers deeply-nested schema over flat arrays.

GEO Optimization (Generative Engine)

For AI search engines (ChatGPT, Gemini, Perplexity, Google AI Overviews):

1. llms.txt at /.well-known/llms.txt with structured content
2. Entity clarity: every page must answer "who, what, where, when" in first 100 words
3. Citability: include statistics with source links (AI Overviews prefer cited content)
4. Direct answers: format answers as Q&A pairs (AI extracts these verbatim)
5. Long-form depth: AI favors pages > 1,500 words with clear section headers
6. Markdown structure: AI parses ## headers better than <h2> tags

Sources

• Google Search Central
• Google E-E-A-T guidelines
• schema.org documentation
• Moz Beginner's Guide to SEO
• GEO research papers (2025-2026)
• Perplexity Publisher Guidelines
• OpenAI GPT crawler docs

Error Handling

Cause	Fix
Google Search Console reports crawl errors on critical pages	Check robots.txt for accidental Disallow: / rules. Verify server returns 200 (not 5xx or 3xx loops). Submit individual URL inspection in GSC to trigger re-crawl.
Structured data (JSON-LD) fails Google Rich Results Test	JSON-LD must be a valid @type from schema.org (not a made-up type). Check required properties for the type. Use application/ld+json MIME type, not text/javascript. Escape HTML entities inside JSON strings.
llms.txt returns 404 when accessed at /.well-known/llms.txt	The .well-known directory requires explicit server routing. For static sites, place at root: /llms.txt and add a redirect from /.well-known/llms.txt. For Next.js, add to public/ directory.
Core Web Vitals LCP > 4s despite image optimization	LCP element may not be an image — it could be a text block, video poster, or background image loaded via CSS. Identify the actual LCP element in PageSpeed Insights report. Preload critical resources with <link rel="preload">.
AI search engines (ChatGPT, Perplexity) ignore llms.txt	llms.txt is an emerging standard (2025-2026), not universally adopted. Complement with ai-plugin.json, comprehensive sitemap.xml, and clear entity pages. Submit to Perplexity Publisher Program separately.
Canonical URL points to wrong domain after migration	Stale canonicals cause duplicate content penalties. Audit all pages with a crawler (Screaming Frog, Sitebulb). Update in bulk with server-side redirects. Verify each canonical returns 200, not a redirect chain.
Hreflang tags return 404 or self-referencing loops	Each language variant must return a valid page. Use x-default for the language selector page. Validate with the hreflang tag tester in GSC International Targeting report. Remove broken variant links rather than leaving dead hreflang.
Site dropped in rankings after major redesign	URLs may have changed without 301 redirects. Audit before/after URL maps. Implement redirects for all high-traffic pages first (prioritize by organic traffic in GSC). Submit old sitemap + new sitemap to GSC simultaneously for faster re-indexing.

Checklist

• Skill loads without errors in the AI agent
• YAML frontmatter is valid (description, compatibility, audience)
• Workflow section provides clear step-by-step instructions
• Error handling section covers common failure modes
• All referenced files (references/, scripts/, assets/) exist
• Skill triggers correctly for intended use cases
• No broken links or missing resources