Building an MCP server for AI-SEO

TL;DR

@automatelab/ai-seo-mcp is a 13-tool MCP server that runs AI-SEO analysis, citation scoring, and structured-data audits directly inside Claude or any MCP-compatible client.

What this is

AI search engines (Perplexity, ChatGPT, Google AI Overviews) pull answers from a small set of sources. Standard SEO tools tell you about backlinks and keyword density - they do not tell you whether a page is structured for AI citation. This MCP server fills that gap: it exposes audit, scoring, rewriting, and crawlability tools that work within your existing Claude context, no browser extension or SaaS dashboard required.

Why we built it

The tooling for AI-SEO was fragmented. You would need one tool to check robots.txt, another to generate llms.txt, another to score citation worthiness - and none of them integrated with the agent you are already using for content work. Claude has no native way to check whether a page will get cited. Wiring it up via MCP was the obvious fix.

Tool surface

Audit

audit_page - full on-page AI-SEO audit: headings, FAQ blocks, entity density, structured data
audit_schema - Schema.org validation and deprecation check
audit_canonical - canonical tag audit and redirect chain inspection

Technical

check_robots - robots.txt parse and AI-crawler directive check (GPTBot, ClaudeBot, PerplexityBot)
check_sitemap - sitemap structure, freshness, and coverage check
check_technical - Core Web Vitals summary and crawlability flags

Scoring

score_ai_overview_eligibility - predicts likelihood of inclusion in Google AI Overviews
score_citation_worthiness - composite score for LLM citation probability

Content

generate_llms_txt - generates a draft llms.txt for a site
validate_llms_txt - validates an existing llms.txt against the spec
extract_entities - pulls named entities and semantic clusters from page content
rewrite_for_aeo - rewrites a passage for Answer Engine Optimization
rewrite_for_geo - rewrites content for Generative Engine Optimization

Install

Add to your claude_desktop_config.json (or equivalent for Cursor / Cline):

{
  "mcpServers": {
    "ai-seo": {
      "command": "npx",
      "args": ["-y", "@automatelab/ai-seo-mcp"]
    }
  }
}

Or run it once to test:

npx -y @automatelab/ai-seo-mcp

Example

After adding the server and asking Claude to audit a page, it runs audit_page and returns something like:

AI-SEO audit: automatelab.tech/n8n-mcp-server/

Citation worthiness: 72/100

FAQ block present: yes (4 questions)

Structured data: HowTo schema detected, no deprecation issues

Entity density: moderate - consider adding more named tools/versions

H2 headings: 6 match common query phrasing

AI-crawler access: GPTBot allowed, PerplexityBot allowed, ClaudeBot allowed

Top recommendation: Add a direct-answer paragraph in the first 150 words. AI Overviews pull from the top of the page first.

That output comes back inside your Claude conversation - no tab-switching, no dashboard login. You can then ask Claude to rewrite_for_aeo the intro block and re-audit in the same session.

What is next

The next version will add per-tool caching so repeated audits on the same URL skip re-fetching, and a batch_audit tool for running a full site check against a sitemap. Feedback welcome - open an issue on the repo or reply here.

Repo: https://github.com/AutomateLab-tech/ai-seo
Landing: https://automatelab.tech/products/mcp/ai-seo/

推荐订阅源

DEV Community