ai-search-blog-optimiser

AI Search Blog Optimiser

Available for both Claude Code and Claude Cowork.

TL;DR

A Claude Code and Claude Cowork plugin for rewrite-only blog optimisation aimed at AI-search citation. It crawls an existing blog with Crawl4AI MCP, builds a site-scoped brand voice baseline, generates evidence-grounded recommendations, and either produces an optimised rewrite or blocks the article truthfully in a local report dashboard.

Give AI Search Blog Optimiser a company blog URL. It indexes the blog, learns and reuses the brand voice, reads Peec MCP data, creates GEO recommendations, and uses a writer agent to turn existing posts into optimised articles.

Most company blogs were written for Google search, not for AI answers, citations, and prompt-shaped buyer questions. AI Search Blog Optimiser helps product marketers, SEO teams, SEO/GEO agencies, and content leads migrate existing owned content forward into GEO.

It turns manual AI-search research and rewriting into a repeatable optimisation loop: review owned content every 2-4 weeks, identify which posts are falling behind competitors or missing from AI answers, and refresh the pages that can win back visibility.

Impact: improve 40, 50, or 100 existing articles with the same evidence-led process so the blog stays current with the category, what is working for the brand, and what is working for the competition.

Quick Start

Claude Code

Install the plugin from the AI Heroes GitHub marketplace inside Claude Code. This is a custom Claude Code marketplace, not the official Anthropic marketplace:

/plugin marketplace add mlobo2012/AI-search-blog-optimiser
/plugin install ai-search-blog-optimiser@ai-heroes-blog-optimiser
/reload-plugins

Then run:

/blog-optimiser https://www.granola.ai/blog --max-articles 3

For targeted article testing:

/blog-optimiser https://www.granola.ai/blog --article-url https://www.granola.ai/blog/sign-in-with-microsoft --article-url https://www.granola.ai/blog/granola-mcp

Run Size And Article Selection

By default, a discovery run processes up to 20 articles from the supplied blog index. This is a default run-size cap, not a hard product ceiling: pass --max-articles N to process a different number of discovered articles.

For targeted work, pass one or more exact article URLs:

/blog-optimiser https://your-blog.com/blog --article-url https://your-blog.com/blog/post-one --article-url https://your-blog.com/blog/post-two

When --article-url is present, the crawler skips index discovery and processes only the supplied URLs in that order. It does not backfill with recent posts, related posts, sitemap URLs, or inferred substitutes.

Crawl Backend

The plugin crawls through Crawl4AI MCP:

• Crawl4AI MCP works with the plugin.
• The local setup uses the official Crawl4AI Docker server exposed at http://localhost:11235/mcp/sse and added to Claude as c4ai-sse.
• Use the c4ai-sse name because the plugin calls the official Crawl4AI md, html, and screenshot tools through that prefix.

On macOS, install Docker Desktop and start Crawl4AI locally:

docker run -d \
  -p 11235:11235 \
  --name crawl4ai \
  --shm-size=1g \
  unclecode/crawl4ai:latest

For Claude Code, add the local Crawl4AI MCP endpoint:

claude mcp add --transport sse c4ai-sse http://localhost:11235/mcp/sse
claude mcp list

For Claude Desktop or Claude Cowork on Mac, use a local stdio bridge in the MCP JSON config so the desktop app can reach the local Crawl4AI server while the plugin runs. Use the local MCP JSON path rather than adding localhost as a cloud remote connector. This pattern requires Node/npm for npx:

{
  "mcpServers": {
    "c4ai-sse": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "http://localhost:11235/mcp/sse"]
    }
  }
}

For your own blog:

/blog-optimiser https://your-company.com/blog --max-articles 10

Claude Cowork

Download the Claude Cowork zip from AI Heroes and install it through the Cowork plugin uploader:

https://www.ai-heroes.co/en-gb/free-tools/ai-search-blog-optimiser

Then run the same /blog-optimiser command from Cowork.

What happens:

• the dashboard opens for the run
• the blog index is crawled
• the plugin pulls article content through Crawl4AI MCP
• brand voice is generated and saved for reuse in later runs
• Peec data is used to find where the brand is missing in AI answers
• competitor and top-cited source patterns are used to shape the recommendations
• recommendations are passed to a writer agent to create an optimised article
• you get markdown, HTML, schema, diff, handoff notes, and a quality manifest

Storage

Writable root selection:

• CLAUDE_PLUGIN_DATA when Cowork provides it
• otherwise platform default roots below
• BLOG_OPTIMISER_DATA_ROOT overrides both for tests/dev only

Default fallback roots: