SEO-GEO-Audit: A Dual-Track Audit Tool Built for the AI Search Era

A Claude Code skill that audits both traditional SEO and Generative Engine Optimization (GEO), helping websites be better understood and cited in AI searches like ChatGPT and Perplexity. Its core value lies in leveraging Claude's capabilities as an LLM to judge whether a website is 'LLM-friendly', filling the gap where traditional SEO tools fail to adapt to the AI search ecosystem.

Traditional SEO has existed for over two decades, focusing on optimizing for crawlers like Google and Bing (meta tags, sitemap, semantic HTML). However, generative AI search tools like ChatGPT and Perplexity are changing the way information is obtained: they read, understand, summarize content, and present answers in a conversational form.

Traditional SEO tools cannot judge whether a website is LLM-friendly, which gave birth to the concept of GEO (Generative Engine Optimization)—an optimization direction designed specifically for LLM consumption habits.

seo-geo-audit is an exclusive skill for Claude Code that can audit both traditional SEO and GEO dimensions of front-end codebases. It uses Claude's capabilities as an LLM to accurately judge a website's LLM-friendliness.

Supports multiple modern front-end frameworks: Next.js (App Router/Pages Router), Astro, Nuxt, SvelteKit, Remix, Gatsby, native Vue, and pure HTML projects. Automatically identifies project types by parsing package.json—no manual path configuration required.

Traditional SEO Audit (30 Items)

Covers metadata completeness (title/description, etc.), semantic HTML, structured data (JSON-LD), basic accessibility, social graph tags, technical SEO (robots.txt/sitemap, etc.).

GEO Audit (37 Items)

Targeted at LLM consumption habits: llms.txt support, content scannability (paragraph length/heading hierarchy), citation friendliness (FAQ/HowTo schema), code presentation (avoid screenshots), server-side rendering, knowledge graph friendliness (entity recognition/relationships).

Core GEO Concept: LLMs need to easily extract key information, understand structure, and find quotable fragments—excessive marketing jargon or information hidden in complex interactions will lower GEO scores.

Output Format

Generates a Markdown report card:

1. Overall Score: SEO/GEO each give an A-F grade + percentage score
2. Problem Classification: High/Medium/Low severity levels
3. Specific Location: Marks file:line for quick fixes
4. Fix Suggestions: Provides code snippet examples

Usage Commands

Command	Function
/audit//seo-geo	Standard audit (SEO+GEO)
/audit deep	Deep audit of all routes
/audit seo	SEO-only audit
/audit geo	GEO-only audit
/audit fix	Generate modification suggestions after audit (requires user confirmation)

Security Design

• Read-only by default, no file modifications
• fix mode previews changes and seeks confirmation
• Only analyzes source code repositories, does not crawl online sites
• Does not modify robots.txt/noai tags/llms.txt without permission

Practical Significance

1. Discover AI search traffic entry points
2. Improve content quality (LLM-friendly = human-friendly)
3. Increase AI citation probability
4. Future adaptation: GEO will be as important as SEO

Technical Highlights

• Framework awareness: Automatically identifies project types
• Modular checks: SEO/GEO maintained independently
• Templated output: Consistent report format
• Ecosystem integration: Part of the Claude Code skill family

Outlook

seo-geo-audit represents a new tool paradigm—serving both humans and AI systems. As AI search becomes popular, GEO will become an independent optimization field, and this tool is an early explorer, providing references for subsequent development.

Front-end developers can get GEO optimization suggestions via simple commands, adapting to AI-era needs without professional background.