A Google Search Console MCP that keeps your AI from misreading the numbers.
Querywright is an SEO MCP server for Claude and ChatGPT. Connect the Search Console property you already own, and your AI can read its finalized clicks, impressions, CTR and average position right in the conversation. What sets it apart is where that data lands: beside measured search demand and your rankings, each labeled for the job it does, so your AI doesn’t mistake one for another.
Setup lives in Settings, and the Search Console guide walks through every step.
Why Search Console data is easy to misread
Search Console reports what Google recorded for your verified site. Handed to an AI on its own, a few of those numbers look like something they aren’t:
- Impressions look like search volume. An impression means Google says your site appeared in Search. It is not the number of people searching that phrase.
- Average position looks like a rank. Google averages every position your site appeared in, across searches, devices and places. It is not one observation of one search.
- A partial period looks like a total. Google finalizes each day a few days after the fact, so the most recent days are often still missing.
- Missing rows look like zero. Google leaves out searches it treats as private or too rare to show, so a day can carry site totals with no search rows at all.
Querywright keeps each of these straight, so your AI reads them for what they are.
Three numbers, three different jobs
Search demand is how much search activity Google Ads reported for a phrase in a market and language you choose.
Rankings are where Querywright observed your site in Google desktop and mobile results for a search you approved, within the top 50, checked weekly.
Search Console is how Google reports your verified site appearing and receiving traffic in Search.
None of them overwrites the others. Querywright never uses an impression count as a demand figure or an average position as a tracked rank, and nothing imported from Search Console changes a plan, approved copy, a check or a tracked ranking.
What your AI can read
Once your property is connected, ask Claude or ChatGPT how the site is doing in Search, which pages Google reports for a search, or whether the import is complete. Your AI reads the same stored evidence you see in Performance, through two tools:
get_search_console_statusshows the connection and import progress.get_search_console_performancereads a date range, optionally narrowed to one search or one page.
These reads use what Querywright has already stored. They don’t contact Google, and they can’t change the connection.
Connecting your property
In Settings, open Google Search Console and pick the Querywright site. Each site links to one Search Console property.
Press Connect Google and approve read-only access on Google’s screen.
Choose the Domain or URL-prefix property that covers your site, then press Use selected property.
You need a Search Console property that already covers the site and that your Google account can open. Querywright then imports up to 90 days of finalized history in the background and keeps adding new finalized days on its own. The Search Console guide covers import progress, reconnecting and removing a connection.
Read-only, by design
Querywright reads Search Console. It doesn’t edit your property, submit URLs, change indexing or modify any setting, and read-only access is the only access it asks Google for. If you remove a connection, your saved Search Console history stays available in Performance and to your AI.
Blank on purpose
When a 30-day window isn’t complete, Querywright shows how many finalized days it has and leaves the summary blank instead of presenting a partial period as a total. A blank isn’t a zero, and it isn’t a failure. If Google returns no search or page rows for a period, Querywright says so rather than inventing detail Google didn’t return.
One record, from plan to results
Search Console is one part of the full Querywright workflow: research the site, measure search demand, approve which page should own each search, write from an evidence-bound brief, check the copy, then track what happens. Each imported day is kept as a dated observation beside your plans, approved copy, checks and rankings, so you or your AI can compare a Search Console trend with the date a page changed.
Search Console records what Google did, not why. Querywright shows the observations with their dates and leaves the conclusion to you.
See how Querywright works with your AI and Rankings for the rest of the record.
Start with Claude or ChatGPT
Querywright is built for freelancers and small agencies already doing SEO work with Claude or ChatGPT. The Google Search Console integration is part of the Founding Beta.
See the Founding Beta offer · Connect Claude · Connect ChatGPT
Frequently asked questions
Does Querywright change anything in my Search Console property?
No. Access is read-only. Querywright doesn’t edit the property, submit URLs, change indexing or modify any Search Console setting.
How much history does it import?
Up to 90 days of finalized history. Google finalizes each day a few days after the fact, so the newest days always arrive a little later.
Does my AI contact Google every time it asks?
No. Claude or ChatGPT reads the Search Console evidence Querywright has already stored, the same data you see in Performance.
Why are some numbers blank?
The 30-day summary stays blank until every day in that window is imported and final, so a partial period is never shown as a complete total. A blank isn’t a zero.
Is an impression the same as search volume?
No. An impression means Google says your site appeared in Search. Querywright keeps search demand as its own number, reported by Google Ads, and never uses an impression count in its place.