Why your average position can improve while your clicks fall

A page-one query that produced nothing

Here are four rows straight out of our own three-month export, taken on 13 August 2026 with no country or page filter applied:

The first row is the one worth sitting with. Position 9.35 is page one. Five hundred and fifty-eight people were shown our result, and not one of them clicked it.

Across the whole property in that period: 8,356 impressions, 14 clicks, a click-through rate of 0.17%, and an average position of 39.8. Ranking was not the problem.

What “average position” is actually averaging

The number is not a rank. It is an average of the highest position your site held each time it appeared, weighted by how often it appeared. That weighting is where the confusion starts, because it means the figure can move without a single page changing rank.

Two ways that happens, both extremely ordinary:

• You lose impressions at bad positions. A page that used to surface at position 60 for a long tail of queries stops surfacing. Nothing improved. The bad rows simply left the calculation, and the average rises.
• You gain impressions at good positions on low-volume terms. A handful of new page-one appearances on queries almost nobody searches will pull the average up while contributing almost no clicks.

In both cases the chart slopes upward and the business sees nothing.

The question that separates the two

Stop looking at the average and sort your queries by impressions instead. Then ask one thing of the top twenty rows:

Is this query getting clicks in proportion to where it ranks?

A query at position 9 should not return zero clicks from 558 impressions. Something between the ranking and the reader is failing, and it is almost always what the result looks like in the results page rather than where it sits.

In our case it was the snippet

When we audited the old site before rebuilding it, we read all 24 pages that carried search traffic. Every one of them was missing its meta description. Not most — all twenty-four.

Those 24 pages carried 50,541 impressions over twelve months. Google was writing every single snippet itself, pulling whatever sentence it judged relevant out of the page. Sometimes that works. Often it produces a line of boilerplate, a fragment of navigation, or a sentence with no reason to click in it.

That is the shape of the problem a rising average position hides: the ranking work had already succeeded. The presentation had never been done.

How to check this on your own site

1. Open Search Console → Performance → Queries, and set the range to three months.
2. Turn on all four columns: clicks, impressions, CTR, position.
3. Sort by impressions, not by position.
4. Flag every row with a position under 15 and a CTR under 1%.
5. For each one, search the query yourself and read your own result as a stranger would.

If the flagged rows have a title that repeats your brand three times, or a description Google invented, you have found it. That is a writing job, not a link-building job, and it is far faster.

If they look fine and still get no clicks, the answer is usually above you on the page — an AI Overview, a pack, or a set of ads pushing position 9 below the fold. That is a different problem and worth knowing you have.