Find the queries on page two
- Open Performance > Search results in Google Search Console.
- Set the date to the last 3 months.
- Switch on Average position.
- Add a table filter on Position, Greater than 10. It takes only one condition, so the cut at 20 comes next (some people go to 30).
- On the Queries tab, filter out your brand, sort by Impressions and skip every row with a position over 20.
The list is never complete: anonymized queries, searched by a few dozen people at most, never show in the table, and exports stop at 1,000 rows.
If you look after a lot of sites, MarketingTracer's Search Insights has this list for every client site under Explore > Keywords. Set Brand to without brand and pick the quick filter Opportunities, which covers positions 4 to 20, a bit wider than page two.
My own site, corewebvitals.io, in MarketingTracer's Search Insights.
Cut the list with four checks
Sorted by impressions, the list looks like a gift, until you see the clicks: 42 on 28,562 impressions, 0.15%. These are four of the biggest rows, plus the one that does earn clicks, and the check that decides each one:
| Query | Impressions | Position | Clicks | Dropped by |
|---|---|---|---|---|
| core web vitals,810,0.13,55,mid | 2,083 | 12.9 | 0 | 1: nobody types it |
| what is cls in core web vitals | 1,315 | 13.6 | 0 | 3: CLS page at 9.2 |
| seo bergen op zoom | 843 | 11.9 | 0 | 2: not what I sell |
| cls web vitals | 802 | 15.1 | 0 | 3: wrong page ranks |
| critical css | 309 | 12 | 14 | Passes: a tool page |
1. Would a person type this?
My biggest row looks like a line from somebody's spreadsheet, and no page on earth should be optimized for it. Read every query: strings of commas and numbers go, and so do instructions like "create a checklist to ensure smooth point cloud streaming performance in browsers."
2. Do you serve it?
Toledo, San Antonio, Stockport and Bern show up in queries from people looking for someone in their own city, and I'm not in any of them. Local SEO in Bergen op Zoom isn't what I sell either. Drop anything whose searcher would never become a client or a reader, whatever its impressions.
3. Does the right page rank?
That 13.6 for "what is cls in core web vitals" is my site's topmost spot, averaged over all impressions. Filter on exactly that query and open the Pages tab: the CLS page sits at 9.2 with 916 impressions, /core-web-vitals at 22.8 with 522. The right page is already on page one.
My own site, corewebvitals.io, in Search Console.
"cls web vitals" goes the other way: the overview page holds it at 11.8 and the CLS page sits at 25.7, so polishing the overview won't help. Two pages fighting over one query is keyword cannibalization. Search Insights' overview flags it in a line that ends with "Pick one."
4. Is there a click to win?
Search the query logged out. An AI Overview, a tool or a row of videos leaves less room for your click. Page two never gave much anyway: in SISTRIX's 2020 mobile data, each position there got under one percent CTR, the share of searchers who click.
And it keeps dropping: on informational queries with an AI Overview, Seer Interactive saw organic CTR fall from 1.76% (June 2024) to 0.61% (September 2025).
My exception is "critical css": 14 clicks on 309 impressions at position 12, 12 of them to my critical CSS generator. People wanted a tool, and the site has one.
Work on the cluster around the page
A query that passes all four checks is worth the work. When a client asks why they're stuck on page two, I always say: let's go and find out. What works for one site doesn't work for another. Apparently your page doesn't have enough weight yet to lift this keyword. Sometimes that's one small change, sometimes a series of small ones, sometimes a couple of big ones. It depends on the reasons behind it.
For the keywords that really matter, I make sure the whole cluster is right. Good content, good links between the pages, and that way the cluster carries the keyword. In practice:
- Wrong page ranks: link to the right one from the pages around it, with the query in the anchor text.
- Right page high on page two: answer the question better and link to it from related pages.
- No click to win: leave it.
Check that it worked
Give it four to eight weeks. Then filter on the query, open the Pages tab and compare the four weeks before and after the change.
Don't stare at average position: it moves slowly and stays an average. Watch the clicks, and which page holds the query.