How Do You Find Which Queries Your Pages Already Rank For?

I want more traffic. Why start with queries I already rank for?

Because those are the only queries where Google has already decided your site is a plausible answer. Getting from position fourteen to position six on a query you already appear for is a far shorter distance than appearing at all for something new. The hard part, which is being considered, is already done.

This walkthrough is for a specific situation. You have a site with some history, probably a blog with a few dozen posts or more, and you want to know what to work on next without guessing. Everything here uses Google Search Console, which is free, and a spreadsheet.

I will go through what the numbers mean, which rows actually matter, and the ones I would deliberately ignore.

What do the four numbers actually mean?

Google defines them precisely, and the precision matters. Clicks are the number of times a user clicked your site from Google Search results. Impressions are how many times your site appeared in Search results. CTR is the click count divided by the impression count. Average position is the average position of the topmost result from your site.

Read that last definition again, because it is the one people misread. Average position is about the topmost result from your site, which means if two of your pages appear for a query, the number describes the better-placed one. That matters when you are trying to work out whether you have one page competing or two.

It is also an average, which means a position of eight might be a page that sits steadily at eight, or one that alternates between three and thirteen. Those are different situations needing different work, and the single number hides which one you have.

Where do you start in the report?

Open the performance report, set the longest date range the report offers you, and look at queries sorted by impressions rather than by clicks. Sorting by clicks shows you what is already working. Sorting by impressions shows you where Google is already putting you in front of people, which is where the unrealized value sits.

Do not change anything else on the first pass. The temptation is to start filtering immediately, and you will understand the shape of your own data better if you look at the unfiltered list first. You are trying to notice surprises, and a filter removes exactly the surprises you have not thought of.

Export the list to a spreadsheet at this point. The in-browser view is fine for looking and poor for thinking, and you will want to sort the same data three or four ways. Google Sheets or Excel is enough. If you later outgrow a spreadsheet, Looker Studio and BigQuery exist for this, and their own documentation is the place to learn them.

Which rows are the actual opportunity?

Three patterns, and the first is the best. Queries with meaningful impressions, an average position somewhere between roughly five and twenty, and a low click count. Those are queries where you are visible but not chosen, and improving the page or the title is often enough to move them.

The second pattern is a query with good position and poor CTR. Being seen at position four and not clicked usually means the title and description are not matching what the searcher wanted, which is a copy problem rather than a content problem and is therefore cheap to fix. I covered that specifically in writing page titles and meta descriptions.

The third is a query you rank for that you never wrote about. Those are gifts, and they are the single most underused thing in this report. Google has told you there is demand you are accidentally serving, which means a page written on purpose would serve it much better.

How do you tie a query to one page?

Filter by the query, then switch to the pages view. That tells you which URL is actually appearing, which is not always the one you assume. On a site with a lot of overlapping content you will sometimes find that the page ranking for your most important query is not the page you wrote for it.

When two of your own pages appear for the same query, you have a decision rather than a bug. Either one page should clearly own it, in which case strengthen that one and point the other at it, or the query is genuinely two different intents, in which case make each page unambiguously about its own.

What I would not do is leave it. Two half-matching pages split the signals and neither becomes the obvious answer, which is the same problem that shows up when answer engines have to choose between them. Deciding which page owns a query is the single highest-value edit in this whole exercise.

What do you do with a query you did not write for?

Write the page you should have written. Use the searcher's exact phrasing as your starting point, because that phrasing is evidence of how people actually ask, and it is more reliable than any phrasing you would invent at a desk. Then link the accidental page to the new one.

Be careful about one thing here. If the accidental page ranks reasonably well, the new page is competing with it unless you make them clearly different. The new page should answer the query directly and the old page should keep its original job, with a link between them rather than overlapping content.

This is where most of my own best-performing posts came from. Not from keyword research in Ahrefs or Semrush, but from noticing that something was already pulling impressions on a phrase I had used almost in passing, and then writing the piece that phrase deserved.

What does a high-impression, low-click row tell you?

Usually one of three things, and they need different fixes. Your result is being shown but is not appealing, which is a title problem. Your result is shown for a query you cannot actually answer, which is a mismatch. Or the answer is appearing on the results page itself and nobody needs to click.

That third case is increasingly common and worth accepting rather than fighting. If somebody asks a question with a one-line answer and that answer appears directly, you were never going to get the click, and chasing it is wasted effort. The right response is to make sure you are the source of that answer rather than to mourn the click, and the same logic now applies when ChatGPT or Perplexity answers instead of sending a visit.

The way to tell these apart is to search the query yourself and look at what comes back. That sounds obvious and almost nobody does it. Thirty seconds of looking at the actual results page explains more than an hour of staring at the numbers.

Which rows should you ignore?

Your own brand name, which tells you about awareness rather than content, and anything with a handful of impressions and no pattern. Long tails of near-identical phrasings are one opportunity, not twenty, so group them by meaning before you count them.

I would also be relaxed about queries where you appear far down and the intent is clearly not yours. Appearing at position fifty for something adjacent to your topic is noise, and trying to improve it means writing content you do not want to own. Not every impression is an invitation.

Also note that some queries may not appear in the report at all, so treat the list as a strong sample rather than a complete census. I would not build a strategy that depends on the list being exhaustive, and the same applies to anything you pull from Bing Webmaster Tools alongside it.

How often should you run this?

Properly, once a quarter. Quickly, once a month. The quarterly pass is the one where you export, group and decide what to write, and it takes a couple of hours. The monthly check is five minutes looking for anything that moved sharply, which is usually either a page you changed or something external.

Resist doing it weekly. The numbers move enough week to week that you will react to noise, and reacting to noise produces edits that undo each other. I have done this and the net effect was a lot of activity and no improvement.

What is worth watching more often is the set of pages you are actively working on. A page you edited last week is a legitimate thing to check, because you have a specific hypothesis about it. The whole-site pass is a different activity with a different cadence, and if your setup is new, start with getting Search Console set up properly.

What should you do next?

Export your queries sorted by impressions, find the five rows with decent impressions and an average position between five and twenty, and look at each one's actual results page. That is an afternoon, and it will give you a better content plan than any keyword tool would.

Then pick the single query you rank for by accident and write that page properly this week. Across 350 published articles, that move has been more reliable for me than anything else in this report, because the demand is already proven rather than assumed. If you also want to see how this intersects with AI citations, I wrote about reading the AI citation side of Search Console.

If you have a Webflow site with real history and want someone to go through this with you and come out with a ranked list of what to write and what to merge, reach out. It is usually the fastest return available on a site that already exists.