Google Search Console Impressions: What One Is Actually Worth

A Google Search Console impression is one appearance of a link to your site on a Google results page that someone was viewing. This is organic search data from the Performance report, not the ad impressions Google Ads reports.

The reader in mind is the founder or first marketing hire who has to explain an impressions chart to someone else, and the worked example is our own property. Between the fortnight ending September 12 and the fortnight ending September 26, 2026, our impressions fell 67.6%, clicks stayed at exactly 6, and average position improved from 73.23 to 61.23. For a tour of every report, see our full guide to Search Console's reports. The question here is narrower: what a single impression is worth, and when a change in the count deserves action.

The short version

• On our site, pages averaging position 51 or worse collected 13,506 impressions in 28 days and not one click. What an impression is worth depends on the position it was earned at.
• Most of our impressions were that kind: 82.6% of query-attributed impressions in the last 28 days sat below position 50.
• Our headline average position improved twelve places (73.23 to 61.23). On the 335 queries present in both periods, it improved about four (78.28 to 74.11).
• Before calling a fall a penalty, check four things: position on the queries you kept, clicks, the device and country split, and the dates on Google's status dashboard.

What counts as an impression in Google Search Console?

In general, a link counts once it sits in the current page of results the searcher is viewing, whether or not they scroll down to it. Google's help page on impressions, position and clicks lists the exceptions: inside a carousel or an expanding result such as an FAQ, the item typically has to be scrolled or expanded into view, and paging away and back within one search does not count twice.

The case that is easy to get backwards is the result Google returned but nobody saw. A link on page 3 records nothing, neither an impression nor a position, unless the searcher actually opens page 3. Being returned for a query is not the same as being seen for it.

Where the link appears	When an impression counts	Position recorded
Standard result on the page being viewed	In general, scrolled into view or not	That result's position
Result on a later page the searcher never opens	Never	None
Item in a carousel or expandable result	Typically once scrolled or expanded into view	Usually the containing element's position
Link inside an AI Overview	Standard impression rules apply	One position, shared by every link in the Overview

The same help page says you should "aim not simply for more impressions, but meaningful impressions", meaning being seen by people who will find the page useful. Our September data shows what the other kind is worth.

How is average position calculated, and why can it move when no ranking does?

It is the topmost position your site held each time it appeared, averaged. Google's worked example: one query shows your site at positions 2, 4 and 6, which counts as 2; another shows it at 3, 5 and 9, which counts as 3; the average is 2.5.

Every impression adds your topmost position on that results page to the average, so the figure leans toward wherever most of your impressions come from. Extend Google's example with a third query where your best result sits at position 90, counting each query once as Google's example does, and the average becomes 31.7. Lose that query again and the figure returns to 2.5, with neither of the first two results having moved.

That is the whole mechanism behind a headline position that improves while nothing ranks better. Deep impressions leave, and the average they were dragging down recovers.

Where you read the number matters too. The Performance report help page says the chart shows the average topmost position for your entire site, while each table row shows the position for that page or query. The chart aggregates by property and the pages table by page, which is why the two totals can disagree: our pages table summed to 16,044 impressions for a 28-day window in which the chart showed 15,938.

What happened to our impressions in September 2026?

They fell in two steps, by two thirds from one fortnight to the next, and clicks did not notice. I pulled the numbers below from the Search Console API on October 2, 2026, for our domain property (sc-domain:growthhasten.com), web search only, with final data through September 29. On the day of the pull, Search Console marked everything from September 30 onward as not yet final, which is expected: the newest days stay preliminary while Google finishes collecting them.

Metric	Aug 30 to Sep 12	Sep 13 to Sep 26
Impressions	15,210	4,930
Clicks	6	6
CTR	0.039%	0.122%
Average position	73.23	61.23

Across the full 28 days from September 2 to 29, the property recorded 15,938 impressions, 15 clicks, a CTR of 0.094% and an average position of 68.68.

The daily series shows two separate drops rather than one break:

Days	Impressions per day (average)	Clicks
Aug 16 to 31	1,512 (peak 2,077 on Aug 30)	19
Sep 1 to 14	919	3
Sep 17 to 29	224	11, six of them on Sep 28

The first step came on September 1: 1,766 impressions on August 31 became 999 the next day, and every day through September 14 stayed between 708 and 1,124. The second came between September 15 and 17: 517 on the 15th, 640 on the 16th, then no day above 327 for the rest of the month. Weighted by impressions, position across September 17 to 29 was 44.98, against 74.59 for September 1 to 14. Clicks were never large enough to carry a trend of their own, which is the first thing to admit about everything that follows.

So what is one impression actually worth?

Below position 50 on our site, nothing. On page one, roughly one click for every fifty impressions. Both readings come from the 28 days ending September 29, with each page grouped by its own average position:

Page's average position	Pages	Impressions	Clicks
1 to 10	47	562	11
11 to 20	12	324	2
21 to 50	28	1,652	2
51 and below	51	13,506	0

Treat this as one small site's 28 days, not a click-through curve. Fifteen clicks is a tiny sample, and they were concentrated: in the second fortnight, 5 of the 6 came from India, and in the final seven days, 8 of 9 did.

They also went mostly to brand pages. The blog's 107 URLs took 15,746 impressions and 2 clicks, while the other 31 URLs (the home page, the app, product, tool and solutions pages, and site pages such as about and privacy) took 298 impressions and 13 clicks. None of the 11 page-one clicks landed on a blog post. So the one-in-fifty figure describes visits to our own brand and product pages, not readers arriving for an article.

The answer is conditional. A page-one impression on a query where people want to visit you is worth something you can measure. An impression on a page averaging position 51 or worse was worth zero here, 13,506 times over.

Did our rankings actually improve twelve places?

A little, and less than the headline says. The property-level average moved twelve places, but the queries that earned impressions in both fortnights moved about four.

Measure	Aug 30 to Sep 12	Sep 13 to Sep 26
Property-level average position	73.23	61.23
Average across named queries only	75.06	69.03
The 335 queries present in both periods	78.28	74.11
Share of impressions in anonymized queries	6.9%	14.0%

The rest of the twelve places is mix. Deep impressions disappeared, which lifts the average by itself, and anonymized queries roughly doubled their share, so more of the total sits in rows nobody can inspect.

Page one did gain ground. Impressions at positions 1 to 10 went from 55 in the first fortnight to 64 in the second while the total fell by two thirds, then reached 102 in the seven days ending September 29 alone.

The daily figure is too noisy to quote on its own at this volume. On September 29 our daily average position was 25.64. Two days later, on preliminary data and roughly 200 impressions a day, it read 55.86. Thirty places in 48 hours, on that volume, is too fast a swing to read as a ranking change.

Why does the queries table show zero clicks when the total shows fifteen?

Because all fifteen came through anonymized queries. Google's deep dive into Performance data filtering and limits defines these as queries not issued by more than a few dozen users over a two-to-three month period, and says they are always omitted from the tables but included in chart totals, unless you filter by query.

On a small site that can be the entire click count. Our queries table attributed 14,498 impressions and 0 clicks to named queries over the 28 days, while the chart showed 15 clicks. The pages table, by contrast, showed all 15.

When the queries table looks empty, read clicks by page. Treat any query filter as a view that leaves the anonymized share out, which on our property was one impression in seven by the second fortnight.

Is a drop in impressions a penalty? Four checks before you decide

Not on its own, and a fall that arrives with steady clicks and better positions on retained queries points away from one. I ran these four checks on our property before treating the September drop as anything more than a change in mix. Each rules something out; none names a cause.

1. Position on the queries you kept: Compare only queries with impressions in both periods. Ours went from 78.28 to 74.11 across 335 queries. A demotion would show up here as those same queries getting worse.
2. Clicks: Ours were 6 in each fortnight. Fewer impressions with unchanged clicks is what you would expect if the impressions that left were producing nothing.
3. Device and country split: Desktop fell 68% and mobile 66%. The United States fell 67% and the United Kingdom 81%. A drop that is proportionate across devices and broad across markets is not a single-surface glitch.
4. Dates against Google's own record: The Google Search Status Dashboard lists the August 2026 spam update from August 18 to 21 and the September 2026 spam update from September 24, still rolling out, with the last core update in May 2026. Our decline came in two steps, on September 1 and between September 15 and 17: the first 11 days after the August update ended, the second a week or more before the September update began. Read the dashboard itself, not coverage of it.

The first check needs volume to mean anything. With only a few dozen queries present in both periods, a four-place move can be a single query changing, so widen the windows before reading it.

The cause is still unknown. The checks ruled out a demotion of retained queries, a loss of clicks, a single device or market, and a dated Google update. For the policy side, here is what the September 2026 spam update does and does not touch.

Which impressions are worth acting on?

The ones at a position that can still produce a click, which on our site meant the top 50. This table is the reading our own numbers support, band by band. On a site with a strong brand or a different query mix, the cut-offs may sit elsewhere.

Position band	What an impression there tells you	What to do
1 to 10	You were seen where people click. A low CTR here is worth fixing, and the snippet is the first thing to check.	Rewrite the title and description, starting with writing a title Google will keep.
11 to 20	Near page one. Our 324 impressions here produced 2 clicks.	Strengthen the page and point relevant internal links at it. This is striking-distance work.
21 to 50	Google returns the page for the query but ranks at least 20 results above it.	Compare it with the results above it for depth and fit to the query, then improve it or merge it into a stronger page.
51 and below	Visibility with no click value at that depth: 13,506 impressions and 0 clicks on our site.	Do not rewrite on the strength of these impressions. Judge the page on its target query, and report them neither as wins nor as losses.

The advice this table narrows is the familiar "high impressions, low CTR, rewrite the title". On page one that is the right fix. At position 85 it aims at the wrong problem: the position is what costs the click there, so the work that matters is whatever moves the page up, not the wording of its snippet.

What should you report instead of impressions?

The numbers that came through our decline intact. Clicks did not move when two thirds of impressions vanished, so they lead the report. Position belongs in the report only for the specific queries each page was written to win, read across weeks rather than days. And page-one impressions were the one impression figure that rose while the total fell, which makes them the version of the metric worth showing.

Raw impressions keep one job: they are the denominator for CTR on page-one rows. Our breakdown of which metrics deserve to be KPIs at an early stage goes further on choosing the rest.

What should you run on your own property this week?

One export. Download the last 28 days of queries from the Performance report with position included, put each row into the four bands from the table above, and read only the 1 to 10 and 11 to 20 rows. Values the report shows as ~ or - export as zeros, so drop any row with a position of 0 before bucketing, or it lands in the page-one band by mistake.

From then on, keep the band split beside the impressions chart whenever you show it to anyone. A total with no position attached tells you how often you were seen, and our September showed how little that can mean.

If you would rather not rebuild that spreadsheet each month, section 02 of GrowthHasten, Search Performance, reads Search Console for you, and the product page lists a "Position spread" view of how your queries distribute across the results and an opportunity type for "impressions without the position to convert them". The figures in this article came from the Search Console API directly, not from the product.