Claude SEO: how to do real SEO work inside Claude

Claude SEO done right: Claude can plan and write SEO but has no search data. How to give it keyword volumes, rankings, backlinks and Search Console data.

Ask Claude for keyword ideas and you will get a tidy, confident list. Ask it how many people search for each one, and it will either guess or tell you it cannot know. Both answers are honest about the same gap: Claude is very good at SEO thinking and has no SEO data.

That gap is the whole story of "Claude SEO". The model can structure a content plan, critique a page, write a title tag that earns the click and spot the logic flaw in a site architecture. What it cannot do on its own is tell you that a keyword gets 900 searches a month in the US, that the top three results are forum threads, that your competitor gained 40 referring domains last quarter or that your page slipped from position 6 to 11 last Tuesday.

This guide shows how to close that gap, what a working setup looks like and which requests get the most out of it.

What Claude does well for SEO without any tools

It is worth being clear about this first, because a lot of SEO work needs no data at all.

Everything on that list depends on judgment. None of it depends on knowing today's search volume.

Where Claude needs real data

The moment a decision depends on numbers, a model without data starts estimating. Estimates are where SEO projects go wrong:

Question Without data With data
Is this keyword worth a page? "It seems popular" 900 searches a month, difficulty 12
Can we rank for it? "Competition may be high" Top results are thin pages from sites with few links
What do competitors rank for? A list of plausible guesses The keywords their pages actually rank for, with positions
Did last month's update help? No idea Clicks and positions from your own Search Console
Are AI assistants recommending us? No idea Saved ChatGPT and Gemini answers that mention you or your competitors

The fix is not a better prompt. It is giving Claude access to the data, so it can look things up while it works instead of guessing.

How Claude gets SEO data: MCP

Claude supports the Model Context Protocol (MCP), an open standard that lets an AI assistant call tools on another service. Connect an MCP server and Claude can use its tools in the middle of a conversation: it decides it needs search volumes, calls the tool, reads the result and keeps reasoning.

SEO Sauna is an MCP server for SEO data. Once connected, Claude can call tools for:

Claude does the talking and reasoning. SEO Sauna supplies the numbers. They are not alternatives to each other; you use them together.

Setting it up

The SEO Sauna MCP server lives at https://app.seosauna.com/mcp.

SEO Sauna Agent setup: copy the setup prompt for Claude Code, ChatGPT or any MCP client, with the MCP server URL https://app.seosauna.com/mcp

Claude Desktop or claude.ai: open Customize → Connectors, choose Add custom connector, paste the URL and approve the sign-in. Custom connectors work on the Free plan too (one connector) and on Pro, Max, Team and Enterprise.

Claude Code:

claude mcp add --transport http --scope user seosauna https://app.seosauna.com/mcp

The first time Claude uses a tool, you sign in to SEO Sauna and approve access. To check that it works, ask:

Use SEO Sauna to check my connection and list my projects.

That read uses no credits. The full guide, including Cursor, Codex and ChatGPT, is in Connect your AI assistant. For step-by-step Claude setup, skills and troubleshooting, see Claude MCP server for SEO, and for how SEO MCP servers compare in general, read SEO MCP server.

Five requests worth trying first

Generic requests like "do SEO for my site" produce generic work. Specific jobs with a clear output work much better. These are good first runs.

1. Save your project context

Set up an SEO Sauna project for example.com. Ask me about the business, who it is for and the main competitors, then save it as project context.

SEO Sauna Context page: business overview, current goal, positioning, writing preferences and competitors that the AI assistant reads first

Every later request reuses this, so Claude stops asking the same questions and its suggestions fit your business instead of a generic one.

2. Find winnable keywords

Find keyword opportunities for [topic] in the US. Keep difficulty under 20, group them by search intent and tell me which three deserve a page first, and why.

SEO Sauna Keyword Research for "running shoes" in the US: 301,000 volume, 147 related keywords, a 12-month trend and the live SERP analysis

Claude pulls real volumes and difficulty scores, then does the part a spreadsheet cannot: judging which keywords match what you actually sell.

3. Study one competitor

Look at competitor.com. Which of their pages bring the most organic traffic, which keywords do they rank for that we don't, and where are they weak?

Domain research works on any site, so this is a quick way to see where a competitor's traffic really comes from.

4. Find striking-distance pages

Using my Search Console data from the last 3 months, find queries where we rank between positions 5 and 20 with decent impressions. Suggest what to change on each page.

This is often the fastest SEO win there is. The pages already rank; they need a better match for the query. Search Console reads use no credits.

5. Check AI visibility

What does ChatGPT answer when someone asks for the best [product category] for [audience]? Which brands does it mention and which sources does it cite?

Search is no longer only ten blue links. Knowing whether AI assistants recommend you, and which pages they cite when they do, is now part of the job.

Reports that stay with the project

When Claude finishes a piece of work, it can save the result as an HTML report in your SEO Sauna project. Anyone in your workspace can open it from the Reports page, print it or export it, and it does not disappear when the chat scrolls away. For agencies this is the difference between a useful chat and a deliverable.

Agent skills: repeatable workflows

MCP gives Claude data. Agent skills tell it how to use that data for a specific job, step by step. SEO Sauna has skills for project setup, keyword research, keyword clustering, competitor analysis, competitive landscape mapping, link prospecting, local SEO, SEO audits, AI prompt research and AI visibility audits.

SEO Sauna Agent setup Skills list: /seo-coach, /seo-audit, /keyword-research, /link-prospecting, /ai-visibility-audit and more, each with a short description

With a skill installed, "run a keyword clustering pass on my saved keywords" follows the same method every time instead of whatever approach the model picks that day. See Agent skills for the list and the downloads.

What it costs

Keyword, competitor, backlink, ranking and AI visibility data comes from paid data providers, mainly DataForSEO, so those lookups use credits. Search Console reads, project context and saved reports use none.

The SEO Sauna plan is €59 a month excluding VAT and includes 17,000 usage credits (€17 worth) each month. You can try it for free first: new accounts get 1,000 trial credits (€1) and no card is needed. Credits are a usage cap, not a fee, so a quiet month simply leaves them unused.

A few habits that make Claude better at SEO

FAQ

Can Claude do SEO on its own?

Claude can reason, plan and write very well for SEO, but on its own it has no live search volumes, rankings, backlink data or Search Console numbers. Connecting an SEO data source through MCP gives it those, so its recommendations rest on real numbers instead of estimates.

Does this work with the free Claude plan?

Yes. Claude's Free plan allows one custom connector, which is enough for SEO Sauna. Pro, Max, Team and Enterprise plans allow more.

Does it work with ChatGPT and Cursor too?

Yes. The same MCP server works with Cursor, Codex, ChatGPT with a custom connection and other MCP clients. See the setup guide for each.

Is my Search Console data safe?

SEO Sauna requests read-only access to Search Console and cannot change anything in your account. Reading your own Search Console data uses no credits.

Do I need to know how MCP works?

No. You paste one URL into Claude's connector settings and approve a sign-in. After that you just talk to Claude as usual.

More from the blog

AI discoverability audit: how to run an AI visibility audit An AI discoverability audit shows whether ChatGPT, Gemini and Google AI name and cite your brand, and why not. Steps, prompts and what to fix first. ChatGPT MCP server for SEO: connect live SEO data Connect an SEO MCP server to ChatGPT as a custom connection on Free or paid plans: step-by-step setup, error fixes and SEO prompts with real keyword data.