Future-proof your SEO strategy by creating content your audience is actually searching for
SEO Simplified will help you identify the words your audience is typing into the search bar, so you can create content that connects with humans, not just search engines.
Subscribe to get your FREE Copy of the SEO SIMPLIFIED keyword research workbook.
Jump To:
Ask Claude what to fix on your website and you might get a decent answer. Ask Claude the same question with your actual search data sitting in front of it, and you’ll get a much better one.
Claude can do real SEO work. It spots patterns in query data faster than I do, and it never gets bored on row 400 of a spreadsheet. But it’s only as good as the data you give it. Without your numbers, it’s working blind from general SEO advice.
By connecting Google Search Console (GSC) to Claude, you’re giving Claude the data it needs.
GSC is the one place Google tells you how it understands your site and how you’re actually performing in search: which queries you turn up for, where you sit, what people click.
Unfortunately, Google doesn’t (yet) offer a native connector to Search Console. There’s no official switch to flip, no button hiding in your settings. If you want Claude reading your search data, you have to get a little creative about how you connect Claude to Google Search Console.
I found 3 ways to do it. One of them I use and give to my students. The other 2 I’d think hard about before handing to a client. But I’ll tell you all about them and let you decide for yourself.
3 ways to connect Claude to Google Search Console
There are three ways to connect Claude to Google Search Console: a third-party MCP, a direct connection to Google’s API, or a connector built through Make.
Option 1. Use a third-party Google Search Console MCP
A third-party Google Search Console MCP connects Claude to your Search Console data through a connector built by an independent developer.
MCP stands for Model Context Protocol, an open standard Anthropic introduced in 2024 for connecting AI tools to outside services. When a service has one, connecting is usually a matter of clicking authorize. Super easy.
Sometimes, an MCP is created by the service itself. In other cases, MCPs are developed by third parties.
In the case of GSC, there is no “official” MCP developed by Google. But if you search “google search console mcp” you’ll find a handful of them, written by independent developers and offered up free.
Here are a few considerations:
Setup: The easiest of the three. Add the connector, sign in with Google, done.
The catch: Google may ask you to authorize the connector to read your Search Console data. That doesn’t give the developer your Google password. It does let their tool use the permissions you approve, so check what the Google authorization screen says.
Plenty of people use independent connectors without incident. I’d still be cautious about giving an unfamiliar provider access to business or client data.
My concern isn’t that the connector gets into everything in your Google account. It’s whether I can evaluate how the provider handles the Search Console access I’ve granted.
Who it’s for: Someone comfortable evaluating the provider and the permissions it requests, especially if that Google account also has access to client properties.
Option 2. Connect Claude directly to the Google Search Console API
A direct Google Search Console connection gives Claude access to your data through Google’s API without putting another service in the middle.
To set this up for myself, I followed Emma Moletto’s guide.
Setup: You create a project in Google Cloud, turn on the Search Console API, create a service account and download its JSON key, then pull her script down from GitHub, install it along with a few Python dependencies, and point a config file at your property. Emma estimates 20 minutes.
It took me considerably longer than 20 minutes.
My holdup was the service account key. Google Cloud refused to create one, because an organization-level security policy was blocking new keys. The confusing part was that the dashboard showed me a newer version of that policy while an older one was still being enforced behind the scenes. I ended up in Google Cloud Shell working out which policy was actually live, switching it off for that one project, and only then could I download the key.
If that made your head spin, me too.
None of that is a knock on Emma’s instructions. It’s what happens when platforms are continuously updating their interfaces and accounts aren’t all starting from the same configurations.
It’s also exactly the kind of afternoon the average business owner doesn’t want.
The catch: It runs on your own machine, through Claude Code or the Claude desktop app. It doesn’t follow you into the Claude web app or into Cowork.
Who it’s for: People comfortable in Google Cloud and a terminal, who want the most secure connection available and don’t mind working for it.
Option 3. Route it through a Make connector
A Make connector gives Claude access to your Search Console data through a connection you authorize inside your Make account.
This is the option I personally use. I’d rather work in the Claude web app than switch over to Claude Code, and I wanted a setup I could hand to my students without asking them to suffer through the Google Cloud part.
Setup: Claude does nearly all of it. You sign in with Google once and confirm which property is yours. The connection and the tool, plus a live test to prove it works, all get built while you watch.
The catch: It needs a Make account. Depending on your usage, the free tier may cover it, or you may need a paid plan (starting at $12/mo).
Make also gets authorized access to your Search Console data. I’m more comfortable making that choice with an established company whose security information, support, and privacy terms I can review. It’s still a third party with access, so check that the Google authorization screen requests read-only Search Console permission.
Before you authorize, you should know that the connection can read every Search Console property that Google account can see.
Who it’s for: Anyone who wants Search Console data in Claude without opening a terminal. It works in the Claude web app and in Cowork, so it follows you to wherever you already work.
Want to set this up yourself?
To set up the Make connection, you need a Make account, the Make connector enabled in Claude, and a Google account with access to your Search Console property.
Then paste this into Claude:
I want to connect you to my Google Search Console data through my Make account, so you can pull my real search performance whenever I ask.
First, ask me which Google account has access to my Search Console property. It may not be the email I use for Make.
Then:
- Find my Make team ID.
- Create a Google Search Console credential request and give me the link to authorize it. Wait for me to confirm I’ve signed in, then verify the connection has read-only Search Console access before moving on.
- Build a Make tool called “GSC Query” that makes an API call to https://www.googleapis.com/webmasters, with inputs for path, method, and a JSON body. Use the google-search-console:makeAnApiCall module and wire in the connection you just created.
- Test it. Run a GET on /v3/sites, show me the list of properties, and let me confirm which one is mine before going further.
- Run one real report: my top 5 queries from the last 28 days, ending 3 days ago.
- Package everything we just set up (team ID, tool ID, connection ID, my confirmed properties, and how to build the requests) into a skill called “gsc-query”, so I can ask for my GSC data in any future chat.
If a call fails twice with the same error, stop and show me the error instead of working around it.
Can I use Zapier instead? Zapier doesn’t have a Google Search Console app (checked September 2026), so this setup needs Make.
Which method should you actually pick?
The right connection method depends on where you use Claude, how technical you want the setup to be, and who you’re comfortable giving access to your Search Console data.
| Setup | Where it runs | What you’re trusting | |
|---|---|---|---|
| Third-party MCP | Minutes | Anywhere Claude runs | An independent developer’s code |
| Direct connect | An afternoon, possibly longer | Claude Code or the Claude desktop app | Google alone |
| Make connector | Claude does most of it | Anywhere Claude runs, Claude web app and Cowork included | Make |
I use the Make connector, for the reason I gave above: I want my search data where I already work, which is the Claude web app.
If you live in Claude Code and you’re comfortable poking around Google Cloud, Emma’s method is the cleanest thing available and it costs you nothing but time. Take it.
What to ask Claude once Search Console is connected
Once Search Console is connected, Claude can analyze your query and page data to find ranking opportunities, performance declines, changes in how Google understands your site, and possible content opportunities.
Here are four useful places to start.
Find striking-distance rankings
Queries where you already rank between positions 11 and 20 are often worth investigating because Google already sees your page as relevant.
Prompt Claude:
Find every query where my average position was between 11 and 20 over the last 3 months. Show me the query, the page Google is ranking, average position, impressions, and clicks. Sort them by impressions so I can see which opportunities have the most search activity.
This gives you a practical list of pages that are already close to page 1. From there, you can look at whether the ranking page could better match the search, answer the topic more completely, or be strengthened in other ways.
Find pages that are losing search visibility
Comparing equivalent date ranges can show you which pages are losing search visibility and which queries are driving the change.
Prompt Claude:
Compare the last 3 months with the previous 3 months. Which pages lost the most impressions or average position? For each one, show me which queries account for the biggest changes.
If your business has enough historical data, you can also compare against the same period last year:
Compare the last 3 months with the same 3 months last year. Which pages lost the most impressions or average position? For each one, show me which queries account for the biggest changes.
The year-over-year version is often more useful for seasonal businesses, but plenty of sites won’t have enough Search Console history yet. In that case, the preceding three months still gives you a useful comparison.
See whether Google understands what you do
Comparing your current query mix with an earlier snapshot can show whether Google is increasingly associating your website with the services, topics, and problems you want to be found for.
Prompt Claude:
Pull my top queries from the most recent 28 days and compare them with a 28-day period from about 3 months earlier.
Group the queries by topic and show me which topics are gaining or losing visibility. Then tell me whether the queries Google is associating with my site are becoming more aligned with the services, topics, and problems I want to be known for. Show me the query data behind your conclusions.
If you have enough historical data, you can run the same comparison against the equivalent 28-day period a year ago.
This is especially useful after you’ve made substantial changes to your website and want to know whether Google is starting to associate your business with the things you actually want to be found for.
Find topics Google is already associating with your site
Clusters of impressions for topics you haven’t intentionally targeted can reveal content opportunities Google is already associating with your website.
Sometimes several related queries start appearing in Search Console even though you don’t have a page specifically focused on that topic. Google is already making the connection. You just haven’t created anything designed to win those searches yet.
Prompt Claude:
Look for clusters of related queries where I’m already getting impressions, but the page Google is ranking doesn’t appear to be specifically focused on that topic.
For each cluster, show me the queries, total impressions, average positions, and the page or pages currently appearing. Then identify topics where Google seems to be associating my site with the subject even though I don’t appear to have a page intentionally targeting it.
Before recommending new content, check whether one of my existing pages would be a better candidate to expand or optimize.
The goal here isn’t to automatically turn every cluster into a new blog post. It’s to find places where Google is already giving you evidence that a topic may be relevant to your site, then decide whether that opportunity deserves a stronger existing page or something new.
The point of all four prompts is the same: let Claude do the tedious part.
It can compare hundreds of rows, group related queries, spot changes, and surface patterns much faster than you can manually. Your job is still to look at what it found and decide what, if anything, is worth acting on.
What Search Console won't tell you
Connecting Search Console to Claude gives you much better information about how your website is performing in Google.
But Google is only one piece of search now.
Search Console can show you when your website appears in Google’s AI features, including AI Overviews and AI Mode. Google added a dedicated Generative AI features report in June 2026.
That report is still pretty limited. You can see how often one of your pages was cited, but not which searches triggered the citation, where you appeared in the response, or whether someone clicked through to your website.
And Search Console can’t tell you anything about your visibility in ChatGPT, Perplexity, Claude, or Copilot. Those are separate platforms with their own data.
So when you connect Search Console to Claude, you’re giving Claude an excellent source of Google search data. You’re not giving it a complete view of your visibility everywhere people might search for you.
There are two other limitations worth knowing about:
Search Console data isn’t real-time. Search Console typically runs 2-3 days behind.. If you ask Claude to analyze your most recent data, I generally recommend ending the date range about three days ago.
The Make connection can’t check whether a specific page is indexed. Google handles that through a separate tool called URL Inspection. If you want to check whether Google has actually indexed a particular page, you’ll still need to do that inside Search Console.
None of this makes the connection less useful. Search Console is still the best source of information you have about how your website is performing in Google. Just be clear about what you’re looking at: Google search data, not your entire search presence.
The connection is only as useful as the questions you ask
Once Claude can access your Search Console data, it becomes very tempting to ask, “What should I do for SEO?” and run with whatever comes back.
I wouldn’t.
The most useful role for Claude here is analyst, not decision-maker.
Ask it to find patterns in your data. Ask it to compare time periods, surface pages that are losing traffic, show you where Google is already giving you impressions, or identify queries where your rankings are close enough to improve.
Then make it show its work.
If Claude recommends updating a page, ask which queries and performance changes led it there. If it says two pages may be competing with each other, ask to see the data behind that conclusion. If it suggests creating something new, check whether your existing site already covers the topic before adding another page.
That’s the part I find most useful about connecting tools like Search Console to AI. Claude can get through a pile of data much faster than I can. But I still want the recommendation tied back to something I can inspect and verify.
You don’t need to hand your SEO over to AI.
You can use AI to make the information you already have far more useful.
Want more practical ways to use AI for SEO?
SEO is changing quickly. So is AI search. And every week seems to bring another tool, tactic, or confident claim about what you’re supposedly meant to be doing now.
I keep a close eye on all of it.
My newsletter is where I share what I’m seeing across SEO and GEO, what’s actually worth paying attention to, and practical ways to use AI tools without handing them more authority than they’ve earned.
If you want to stay current without making this your full-time job, join the newsletter here.
