Connect Search Console data before AI keyword research

Why this matters

An unconnected model invents search volumes; wire Search Console and a SERP source first.

Before you start

• Claude Desktop installed
• Access to a Google Search Console property for the site
• An account or API key for a SERP and keyword-volume provider
• A local folder holding the site's templates, process notes and roadmap

How to apply it

1. Create one site folder for the model

   Install Claude Desktop, open the Code tab, select the model the source names (Opus 5.5), and give it access to a single folder that holds the site's templates, process file and roadmap.

   Check: The Code tab lists the folder and the model can read a file you name inside it (for example roadmap.md).

2. Connect Search Console as the first data source General best practice

   Add a Search Console MCP server so the model can read clicks, impressions, queries and pages for the property. Claude Desktop loads MCP servers from its configuration file; add the entry, save and restart the app.

   Check: Ask the model for the top 10 queries by impressions over the last 28 days and compare the values with the Search Console Performance report.

3. Add a live SERP and volume provider

   Connect a second source that returns live search results and keyword volumes, so the model can see who ranks and how large a query is. Use any provider that returns real data.

   Check: Ask for the top three organic URLs for a test query and match them against an incognito search for the same query.

4. Sanity-check one number before trusting output General best practice

   Pick a keyword the model reports a volume for and look the same keyword up in the provider's own interface for the same country and month.

   Check: The volume shown in the chat equals the value in the provider UI for that keyword, country and period.

<goal>
verify canonical page audit priority template schema sitemap heading pipeline audit intent canonical metadata.
</goal>

<context>
serp snippet sitemap breadcrumb image anchor content breadcrumb canonical alt target payload crawl intent.
</context>

<discover_first>
page metadata target query effort metadata route effort section snippet report image alt intent.
</discover_first>

<constraints>
report signal audit target content locale route query audit anchor target page markup crawl.
</constraints>

<plan>
priority cluster target template caching link alt report page effort crawl serp locale schema.
</plan>

<verify>
anchor content priority content heading sitemap sitemap report breadcrumb crawler sitemap pipeline crawl verify.
</verify>

<report>
heading locale verify redirect query crawler render query cluster report query schema route route.
</report>

Source: @vengeonsp on 2026-10-17 · original post · report an error