How we replaced SEMrush with DataForSEO + Claude Code.

What you'll get.

Two tools — DataForSEO's API (via MCP) and Claude Code — running in a terminal. Claude pulls the data, analyzes it, builds the report, creates tasks in Notion, and deploys the whole thing. Your job is to ask the right questions and review.

• Run a full SEO audit — keywords, competitors, on-page, AI search visibility — from your terminal
• Generate interactive HTML reports automatically
• Query live SERP data, keyword volumes, and competitor metrics on demand
• Test how your brand appears across ChatGPT, Claude, Gemini, and Perplexity (GEO audit)
• Push action items directly into a Notion task database
• Do all of this for ~$3–6 per audit instead of $130–500/month per tool seat

What you need.

• Claude Code — Anthropic's CLI tool (install guide)
• DataForSEO account — pay-per-use API, no subscription (dataforseo.com)
• Node.js — for the MCP server connection
• Notion account — optional, for task management integration
• A terminal — Mac, Linux, or Windows with WSL

The setup.

npm install -g @anthropic-ai/claude-code

{
  "mcpServers": {
    "dataforseo": {
      "command": "npx",
      "args": [
        "-y",
        "dataforseo-mcp-server"
      ],
      "env": {
        "DATAFORSEO_USERNAME": "your-email@example.com",
        "DATAFORSEO_PASSWORD": "your-api-key-here"
      }
    }
  }
}

notion.so/your-workspace/abc123def456...?v=...

Run your first audit.

Start Claude Code in your project directory:

claude

Now just ask for what you need in plain language.

Keyword research

Claude calls dataforseo_labs_google_ranked_keywords and returns a formatted table with keywords, volumes, positions, and ranking URLs.

Competitor analysis

Claude calls dataforseo_labs_google_competitors_domain and gives you shared keywords, ETV, and competitive positioning.

On-page crawl

Claude calls on_page_instant_pages for a technical snapshot: TTFB, DOM size, content rate, missing elements, and recommendations.

AI search visibility (GEO audit)

This is something no traditional SEO tool offers:

Claude calls ai_optimization_llm_response for each model and compares: is the brand mentioned? Cited as a source? What position? You get a cross-platform visibility matrix.

Generate a report.

Claude writes a complete, self-contained HTML file with CSS, tables, and navigation. No dependencies — just open it in a browser.

You can deploy it instantly:

Claude creates the repo, pushes the file, enables Pages, and gives you a shareable URL. From "audit this site" to "here's a live report" in 10–15 minutes.

Push action items to Notion.

This is where the audit becomes operational. Instead of a PDF that sits in someone's inbox, the findings become trackable tasks:

Claude creates a Notion page for each action item with:

• Clear task title (e.g. "Write /tips-rad/ article for 'tåliga perenner' — 480 sök/mån")
• Priority based on impact (search volume × competitive gap)
• Category tag (SEO, Content, Technical, GEO)
• Detailed notes with Claude's analysis and specific recommendations
• Source reference back to the audit finding

Full audit in 5 prompts.

Here's the exact prompt chain for a complete audit:

1. Pull the top 50 ranked keywords for [domain] in Sweden. Show volume, position, and URL.
2. Who are their top 10 organic competitors? Show keyword overlap.
3. Crawl the homepage and top 3 landing pages. Check technical SEO.
4. Test these 4 real search queries against ChatGPT, Claude, Gemini, and Perplexity: [queries from step 1]
5. Create an HTML report, deploy to GitHub Pages, and push the top 10 actions to our Notion database.

Total time: ~15 minutes. Total cost: ~$3–6.

What it costs.

Real cost breakdown from a recent audit of blomsterlandet.se:

You'd need to run 700+ audits per year to match the cost of one SEMrush seat.

Better & worse than traditional tools.

What's better

What's worse

Who this is for.

• Agencies — doing SEO audits for multiple clients. Cost savings compound fast.
• In-house SEO teams — with technical capacity who want deeper analysis.
• Consultants — who want sophisticated reports without enterprise tooling.
• Anyone curious about GEO — this is the easiest way to test it.

Not for: teams that need point-and-click dashboards, anyone without terminal comfort, or use cases where historical rank tracking is the primary need.

Getting started.

1. Install Claude Code — npm install -g @anthropic-ai/claude-code
2. Create a DataForSEO account and get your API credentials.
3. Create .mcp.json in your project folder with the config above.
4. Optional — set up Notion integration for task management.
5. Run claude and ask: "Pull the top 20 keywords for [your-domain]".
6. Ask Claude to build an HTML report and push tasks to Notion.