GEO / AI SEARCH
OPTIMIZATION

When you run /seo geo, Claude SEO reviews 5 weighted dimensions: citability, structural readability, useful media, authority signals, and technical accessibility. The resulting GEO Readiness Score is an internal diagnostic rubric, not a predicted citation rate or a guarantee of visibility.

Google's current guidance is straightforward: established SEO practices also apply to its AI features. Helpful, reliable, crawlable content remains the foundation. Claude SEO uses GEO checks to organize that review across Google AI features, ChatGPT, Perplexity, and Bing Copilot.

Primary references: Google's generative AI optimization guide explains that established SEO practices also apply to AI features, while its AI features documentation covers eligibility, preview controls, and traffic measurement.

What gets analyzed

The GEO analysis evaluates 5 practical dimensions that can be inspected on a page. Third-party correlation studies can inform hypotheses, but they do not prove that a specific mention, passage length, or content format causes an AI citation. Recommendations therefore stay evidence-led and platform-specific.

The GEO Readiness Score is a consistent internal review framework across Google AI features, ChatGPT, Perplexity, and Bing Copilot. It summarizes observable page signals and should be interpreted with the report's evidence and limitations.

Step 1: Install Claude SEO

git clone --depth 1 https://github.com/AgriciDaniel/claude-seo.git
bash claude-seo/install.sh

Step 2: Open Claude Code

claude

Step 3: Run the GEO analysis

/seo geo https://your-site.com

The analysis reviews AI crawler access, evidence and answer structure, optional llms.txt documentation, and platform-specific controls. Its readiness score is a vendor checklist, not a platform signal or citation forecast.

The base skill can make a qualitative evidence review from accessible public pages. Observed ChatGPT responses or cross-platform LLM mention data require the optional paid DataForSEO extension, provider credentials, and cost approval.

Citability is only half of the AI question. For agents that browse, fill forms, and act for people, the separate agent readiness audit (/seo agentic, added in v2.4.0) reports the Lighthouse Agentic Browsing result and checks access policy, llms.txt, discovery files, and WebMCP. Full audits run both skills side by side.

Crawler-purpose boundary

The v2.2.5 skill source checks GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, and PerplexityBot, then recommends several together for AI search. That table is not purpose-current. For search discovery, review OAI-SearchBot, Claude-SearchBot, and PerplexityBot. GPTBot and ClaudeBot are training controls. ChatGPT-User, Claude-User, and Perplexity-User are user-request agents with different robots behavior. Recheck the official OpenAI, Anthropic, and Perplexity documentation before acting on the generated robots.txt recommendation.

Quick wins the analysis identifies

• Answer the page's primary user need clearly and early
• Make important passages self-contained without forcing a target word count
• Use descriptive H2 and H3 headings that reflect the actual section content
• Support material claims with relevant primary sources
• Review search access for OAI-SearchBot, Claude-SearchBot, and PerplexityBot, separately from training and user-request agents
• Treat /llms.txt as optional documentation for services that choose to read it, not as a Google ranking signal
• Implement Person schema for authors with sameAs links to external profiles