Find striking-distance keywords with AI

Striking-distance keywords are queries where Google already ranks one of your pages in positions 6 to 20, so a focused edit can lift it onto page one. Paste the prompt below into Claude or ChatGPT and you get a ranked list of those queries, the page behind each one, and the first pages to improve.

The prompt

Paste it into Claude, ChatGPT or any client connected to the Google Search Console MCP server. Replace [your property] with your site.

Using Search Console data for [your property], find my striking-distance keywords: queries where I rank between position 6 and 20 over the last 28 days, with at least 10 impressions. 1. List each query with its ranking page, average position, impressions and clicks, sorted by impressions. 2. Group the queries by page, so I can see which single page has the most striking-distance demand behind it. 3. For the top 3 pages, pull every query that page ranks for (not just the 6 to 20 ones) so you can see what the page is actually being matched to. 4. For each of those 3 pages, tell me the one change most likely to move it onto page one: a title or H1 that matches the query wording, a new section that answers the query, or internal links from stronger pages. Finish with a priority order (page, why, the change) and skip anything with fewer than 10 impressions unless the list would be empty.

What the agent does

Tools it calls, in order: search_insights, search_analytics. About 2 to 4 data pulls per run (what counts as a data pull).

  1. Shortlist the queries. The agent calls search_insights with the striking_distance recipe. That returns queries at average positions 6 to 20 with their ranking page, impressions and clicks, already sorted by impressions. One call, one data pull.
  2. Group by page. No extra call: the agent regroups the shortlist so a page with five near-miss queries rises above a page with one.
  3. Read each top page's full query set. For the top pages, the agent calls search_analytics with dimensions query,page and a page filter. This shows the queries ranking well, the ones in striking distance and the ones far below, which tells you what Google thinks the page is about. One data pull per page, so one to three here.
  4. Recommend one change per page. The agent compares the query wording with what it knows of the page (or with the page itself, if your client can fetch URLs) and names the edit most likely to help.

A full run is about 2 to 4 data pulls, which fits inside the free plan's 15 a day. If you only want the list without the per-page breakdown, stop after step 1 and it costs one.

Example output

This is a real run on mcpsearchconsole.com, covering 2 September to 30 September 2026 (the last 28 days of final data at the time). The site is young, so the noise floor was lowered to 3 impressions to show more than one row. Every row below came back from search_insights; nothing was added by hand.

QueryRanking pageAvg positionImpressionsClicks
mcp search console/8.8130
google search console mcp for claude/8.360
search console mcp claude/8.060
google search console claude mcp/9.850
mcp google search console claude/9.030

Two things stand out. First, every striking-distance query lands on the homepage, so grouping by page gives one clear target. Second, four of the five queries pair "Search Console MCP" with "Claude". The site has a dedicated guide for that pairing, but in this window Google matched these queries to the homepage instead, at positions 8 to 10. The follow-up search_analytics call on the homepage showed the same pattern at larger scale: "google search console mcp" had 136 impressions at position 54.0, so the broad head term is still far off while the Claude variants are close.

The recommendation the agent gave: make the Claude angle explicit on the homepage (a heading and a short section on using it with Claude) and add a prominent link from the homepage to the Claude guide, so whichever page Google prefers has the words it needs.

How to act on it

Each row falls into one of a few cases. Use the rule that matches.

  • Position 6 to 10, decent impressions, zero or few clicks. You are on page one but low. Rewrite the title so it starts with the query's wording and check the description answers it. This overlaps with fixing low-CTR pages; do that first because it is the cheapest edit.
  • Position 11 to 20. You are on page two. Titles alone rarely move this. Add a section that answers the query directly (an H2 phrased like the query, then the answer in the first two sentences under it), and add two or three internal links to the page from pages that already get clicks.
  • Several queries on one page. Work on that page first. One edit can lift every query in the group, which is why the prompt groups by page.
  • The ranking page is the wrong page. If the query belongs on a different page you own (as with the Claude queries above landing on the homepage), strengthen the right page and link to it from the one Google picked. If both pages keep trading places, treat it as keyword cannibalization instead.
  • Very low impressions (under 10 in 28 days). Note it and move on. The position average is too unstable to act on, and the click upside is small even if it reaches the top three.

Change one page at a time and give it two to four weeks before judging. Search Console data trails by about two days and positions move slowly, so ask the agent to re-run the same prompt next month and compare.

Variations

Try this promptRun the striking-distance check for [your property] but only for pages under /blog/, and exclude any query that contains my brand name.
Try this promptFor [your property], list striking-distance queries at positions 11 to 20 only, then tell me which page should get internal links pointing to it and from which of my top 5 pages by clicks.
Try this promptRepeat last month's striking-distance check for [your property] and compare: which queries moved onto page one, which slipped further, and which are new on the list?

Limits

  • Average position is an average. Position 9 can mean you rank 5th on mobile and 14th on desktop, or 3rd in one country and 20th in another. Ask the agent to split by device or country before rewriting a page for a query that averages out.
  • No search volume or difficulty. Impressions are the only demand signal, and they only count searches where your page appeared. A query at position 18 with 12 impressions may have far more volume than that; Search Console cannot tell you how much.
  • No competitor view. The data says where you rank, not who outranks you or why. Look at the live results page before deciding what the new section should say.
  • Anonymized queries are missing. Google withholds rare queries for privacy, so a page can show impressions with few or no query rows behind them.
  • Fresh data lags. The last two days or so are not final, so the window ends a little before today.

Comparing tools for this? See our round-up of the best SEO MCP servers. To set up the connection first, follow the guide to use Search Console in Claude.

Questions

Related use cases

Run this on your own site in about a minute.

Connect Search Console free