# Google Search Console MCP: how to connect it, what it can see, and the questions worth asking > There's no official Google Search Console MCP. Here's how to connect GSC to Claude or ChatGPT, the API limits to know, and the prompts worth running. Source: https://contengi.com/blog/google-search-console-mcp Author: Stefan Maritz Published: 2026-11-06 Search Console is the bank statement of SEO, and an MCP server lets your AI read it for you. **TL;DR** - A Google Search Console MCP server connects your Search Console data to an AI assistant such as Claude or ChatGPT, so it can query clicks, impressions, CTR, position, and index status in plain English. - Google doesn't publish an official Search Console MCP server, so every option is third-party code running on the free Search Console API. - You have three routes: self-host a community server with your own Google Cloud credentials, use a hosted connector, or use a platform that already has Search Console wired in. - The API limits shape every question: 1,200 search analytics queries per minute per site, 2,000 URL inspections per site per day, data that is typically two to three days old, and up to 50,000 rows a day per search type. - The prompts that pay off are striking-distance keywords, pages losing clicks, queries with no matching page, and good rankings with weak click-through. Search Console is the bank statement of SEO. It arrives a few days late, nobody reads it for fun, and it's the only record of what Google did with your site. An MCP server lets an AI read the statement for you and point at where the money went. This post covers how to connect Search Console to Claude or ChatGPT, what the API will and won't hand over, and the questions we ask it every week. For the rest of the stack, start with the guide to [SEO MCP servers worth connecting in 2026](https://contengi.com/blog/seo-mcp). ## What is a Google Search Console MCP server? A Google Search Console MCP server is a connector that exposes the Search Console API to an AI assistant through the Model Context Protocol. The assistant gets tools for search analytics, sitemaps, and URL inspection, and calls them when you ask about your search performance. Search Console is missing from [Google's official list of MCP servers](https://github.com/google/mcp), even though Google Analytics is on it. So someone else builds every Search Console MCP, on top of the [Search Console API, which is free of charge](https://developers.google.com/webmaster-tools/pricing) within its usage limits. ## How do you connect Search Console to Claude or ChatGPT? Pick your route by how much Google Cloud you want in your life. Route What you do Good for Watch out for Self-host a community server Create a Google Cloud project, enable the Search Console API, create credentials, and run the server locally Developers who want it free and under their control You maintain it and you guard the credentials Hosted connector Paste a URL into your AI client and sign in with Google Marketers who want it working today A subscription, and a third party holding your token Platform with Search Console built in Connect Google once inside the platform, then reach the data through its MCP Teams that want the data next to briefs, drafts, and publishing You get the tools that platform chose to expose ### Self-hosting a community server The open-source route is free and fiddly. A server like [mcp-server-gsc runs through npx with a Google credentials file](https://github.com/ahonn/mcp-server-gsc), so you need a Google Cloud project, the Search Console API switched on, and a service account or OAuth client with access to your property. Budget an hour the first time, and add the service account's email as a user on the property, or every query comes back empty. ### Hosted connectors and platforms Hosted connectors skip Google Cloud and handle the sign-in for you. Platforms go a step further and put Search Console next to the work it informs. We use Contengi's own MCP connection, which pulled the queries in this post, and we draft and refresh posts from the same chat. If you'd rather stay in a dashboard, there are [AI tools for Google Search Console analysis](https://contengi.com/blog/ai-tools-for-google-search-console-analysis) that skip MCP entirely. ## What can a Search Console MCP see, and what are the limits? It sees what the API exposes: queries, pages, countries, devices, and search appearance, each with clicks, impressions, CTR, and average position. It can also inspect single URLs for index status. The limits decide what a sensible question looks like. - **Speed:** search analytics allows [1,200 queries per minute per site and per user, and URL inspection allows 2,000 calls per site per day](https://developers.google.com/webmaster-tools/limits). - **Freshness:** [data is typically available after two to three days](https://developers.google.com/webmaster-tools/v1/how-tos/all-your-data), so anything from the last couple of days are incomplete. - **Volume:** search analytics returns up to 50,000 rows a day per search type, sorted by clicks, so the long tail of zero-click queries falls off first. - **Detail against completeness:** grouping by page or query drops some data to keep the request fast, so query rows never add up to the site total. - **Cost of a question:** long date ranges and page-plus-query grouping burn the most quota, so ask for 28 days by page before you ask for a year by everything. The completeness point is the easy one to miss. Contengi's own Search Console tool warns about it on every query-grouped report, because an assistant that adds up query rows will report a drop that never happened. When you want a total, ask for totals with no dimensions. ## Which questions should you ask a Search Console MCP? The useful questions compare things the dashboard makes you click between. 1. **Striking-distance keywords.** "Show queries where we rank between positions 8 and 20 with at least 100 impressions in the last 28 days, and the page that ranks for each." These are the cheapest wins on your site. 2. **Pages losing clicks.** "Compare clicks by page for the last 28 days against the 28 days before, and list the 10 biggest drops." That's your [content decay](https://contengi.com/blog/content-decay) list, ready for a refresh. 3. **Queries with no matching page.** "Which queries get impressions but land on a page about something else?" Those are the posts you haven't written yet. 4. **Good position, weak click-through.** "Find pages in the top five with a CTR under 2%." The ranking is doing its job, so rewrite the title and meta description. 5. **Index checks after publishing.** "Inspect these five new URLs and tell me which are indexed." It beats pasting them into the inspection tool one at a time. When we ran a query filter for "mcp" on contengi.com while writing this, every result was a full sentence, such as "which headless cms platforms have mcp connectors in 2026?". That's a prompt typed into Google, and it's why each section here answers its question in the first line. ## Checklist: get Search Console into your AI this week 1. **Choose your route.** Self-host if you like Google Cloud, hosted if you don't, or a platform if you want the data next to your drafts. 2. **Grant read-only access first.** Analysis only needs read access, so start there and add more if a tool asks for it. 3. **Check the connection with a totals query.** Compare the AI's total clicks for last month with the Search Console dashboard before you trust anything else it says. 4. **Save your five prompts.** Keep the questions above in a note and run the same checks every week. 5. **Act on one finding a week.** Refresh one decaying page or rewrite one weak title, check it again in four weeks, and fold the results into your next [content audit](https://contengi.com/blog/content-audit). ## Frequently asked questions ### What is a Google Search Console MCP? It's a connector that lets an AI assistant query your Search Console data through the Model Context Protocol. You ask questions in plain English and it pulls clicks, impressions, CTR, position, and index status from the Search Console API. ### Is there an official Google Search Console MCP server? No. Google's official MCP list includes Google Analytics but not Search Console, so every option is a community or commercial server built on the Search Console API. ### Is a Search Console MCP free? The Search Console API is free within its usage limits, and community servers are free to run yourself. Hosted connectors and platforms charge for the convenience of skipping the Google Cloud setup. ### Can ChatGPT connect to Google Search Console? Yes, through an MCP server. Add a remote Search Console MCP as a custom connector in ChatGPT's developer settings, or use a hosted service that supports ChatGPT directly. ### Can a Search Console MCP request indexing? URL inspection through the API reports whether a URL is indexed, but it doesn't submit the URL for indexing. You still request indexing inside Search Console itself.