Search Console Impressions But No Clicks

Why does Search Console show impressions but no clicks?

Because an impression is a far lower bar than a click, and most impressions are recorded in places nobody is looking. Search Console counts the appearance; the click needs a human to read the result and choose it.

It is also the normal condition now rather than a fault. Rand Fishkin's SparkToro (opens in a new tab) analysis of Similarweb's US clickstream panel found that 68.01 per cent of Google searches ended without a click in the first four months of 2026, against 60.45 per cent in 2024. Most searches now end where they started.

Five causes account for nearly all of it. They are not equally common and they do not respond to the same fixes, which is why a single list of tips rarely helps.

1. The page ranks too deep to be read. This is the largest cause on most sites and the one people spend least time on, because it looks like a content problem and is a position problem.
2. An AI surface answered the query. AI Overviews and AI Mode generate impressions, and a reader who gets the answer in place has no reason to click.
3. The search intent does not match what the page offers. Google shows the page, the searcher reads the title, and the title tells them this is the wrong thing. Search intent mismatch produces high impressions and low clicks even on page one.
4. A result type that rarely earns clicks. Image packs, People Also Ask panels, sitelinks and knowledge panels all log impressions against your site without ever showing a plain blue link.
5. A reporting artefact. The impressions were logged by Search Console but never seen by a person, or never happened as recorded.

The fifth cause has grown enough since May 2025 to deserve its own section, and it is the one missing from almost every guide on this subject.

What counts as an impression in Search Console?

Google counts an impression when a user has seen, or could have seen, a link to your site. The second half of that definition does the damage. Google states the rule plainly in its documentation on impressions, position and clicks (opens in a new tab): for standard results, an impression is counted whenever an item appears in the current page of results, whether or not the item is scrolled into view.

Within one page of results that clause matters: positions five to ten collect impressions from people who stopped reading at result four. Depth works differently. Google returns ten results to a page, so a listing at position 67 records an impression only when somebody loads the seventh page of results, and in practice that is an automated tool rather than a reader.

Carousels and expandable sections have their own rule. Google says the item must typically be scrolled into view within the carousel, or expanded by a click, before an impression registers, so carousel impressions track attention more closely than ordinary blue links do. Images in the main combined results are not an exception to anything: Google counts those whether or not they are scrolled into view.

The click definition has a matching trap. Google counts a click only when it sends the user to a page outside Search, Discover or News, and a link that keeps them inside Google is not a click. Expanding a People Also Ask box or opening an image in the panel leaves your impression count higher and your click count untouched.

AI Overviews are a common exception to get wrong, so it is worth stating. Google's documentation says standard impression rules apply to both AI Overviews and AI Mode, and that clicking a link to an external page in either counts as a click. AI surfaces suppress clicks by answering the question, not by logging a kind of impression that cannot convert.

What changed in Search Console impression reporting since May 2025?

Three changes to Google Search Console, and together they mean impression counts before and after them are not comparable. Every widely read guide to this question predates all three, which is why the usual advice skips straight to your title tag.

The first of those is the one people miss. Google began counting AI Mode in the ordinary Web search type in June 2025, so impressions from a surface that had never been reported started arriving inside the same total as blue links, with nothing in the default view marking them.

The third reset the baseline on most sites. Rank-tracking tools had been pulling 100 results in a single request, and every one of those results logged an impression for the site it named. When Google stopped supporting the parameter in September 2025, Tyler Gargula's analysis of 319 properties found 87.7 per cent saw impressions fall and 77.6 per cent lost unique ranking keywords, with short and mid-tail terms hit hardest. No traffic changed. The scraped impressions simply stopped being counted.

The logging error is the one worth knowing about, because it ran for 349 days. Google's own data anomalies page (opens in a new tab) records it plainly: a logging error prevented Search Console from accurately reporting impressions from 13 May 2025 until 27 April 2026. Clicks were not affected, so every site's CTR for that period reads lower than it really was.

Whether the historical figures were ever corrected is not stated. Google says the issue has been resolved and records no restoration of the affected period, which is worth contrasting with the August 2026 entry on the same page, where Google does say the missing data was restored. Treat anything before 27 April 2026 as a period you cannot compare against.

That page is worth reading in full before you interpret any impression trend. Google logged seven separate dated entries in 2026 alone describing impression reporting errors, covering ordinary Search results, Discover, the generative AI report in Search, and job listings. Reporting errors in this metric are routine rather than exceptional.

One change often listed alongside these is not one. The generative AI performance report launched on 3 June 2026 and reached all sites on 31 August 2026, and it reports impressions only, with no clicks, CTR, queries or position.

It counts nothing new. Google states that the report draws on data already in the Web search type, so it is a separate view rather than an addition. Use it to see how much of your total comes from AI surfaces, and read our comparison of Google AI Mode and AI Overviews for what it will and will not separate.

Which of your impressions could ever become clicks?

Sort them by average position before anything else, because position decides whether any on-page change can help at all. Export the Queries tab with the Average position column, group the rows into bands, and read each band this way:

Query-level rows in Search Console are sampled and anonymised, so the bands you build will not sum to the site total. Read them as a large sample rather than a complete decomposition.

The practical rule that falls out of it: impressions below about position 20 are a record that Google considered you, not an audience you failed to convert. Counting them in a CTR figure drags the number down and tells you nothing you can act on. If the positions themselves have fallen, that is a different diagnosis: why Google rankings drop and what to check first.

Compare each band against a benchmark for that band, never against a sitewide average. Backlinko's analysis of four million results puts position one at 27.6 per cent, with the top three taking 54.4 per cent of all clicks. Advanced Web Ranking republishes the same benchmark monthly from Search Console data, split by whether an AI Overview is present, which is the version worth bookmarking. Those are the numbers your top-three rows answer to. Your position-60 rows answer to nothing.

Is the 61 per cent AI Overviews CTR drop still the right number?

Not as most people quote it, and the omission is bigger than the staleness. The figure comes from Seer Interactive and measures something narrower than the shorthand suggests: organic CTR on informational queries showing an AI Overview fell from 1.76 per cent to 0.61 per cent between June 2024 and September 2025, across 3,119 terms and 42 organisations.

The number almost nobody repeats sits in the same study. Over that same window, CTR on queries with no AI Overview at all fell from 2.72 per cent to 1.62 per cent, a drop of 41 per cent. Whatever was removing clicks was not confined to pages with an AI Overview on them, so attributing the full 61 per cent to AI Overviews overstates the case by a wide margin.

Seer's 2026 update (opens in a new tab) runs on a far larger sample: 53 brands, 5.47 million queries and 2.43 billion organic impressions from January 2025 to February 2026. It puts organic CTR on AI Overview queries at 2.36 per cent in February 2026, recovered from 1.31 per cent in December 2025, against 3.82 per cent where no AI Overview appears. That is a gap of 38 per cent between the two, on a sample roughly 1,750 times the size of the one behind the famous figure.

Seer also documents the arithmetic this article is about. In the segment where brand-cited CTR fell 61 per cent between the third and fourth quarters of 2025, impressions ran from 15.8 million to 33.1 million while clicks moved from 398,798 to 400,271. Clicks were flat. The denominator doubled. Seer's own recommendation, stated twice in that update, is to measure impressions and clicks separately rather than reading CTR as a performance signal.

A falling CTR on a page whose clicks are steady is arithmetic, not a decline. That distinction decides whether you have a problem at all.

The pattern has a name in the industry: the great decoupling, meaning impressions in Google Search Console climbing steadily while clicks stay flat. Naming it helps, provided you remember that the curve has at least four causes. AI surfaces are one. A logging error, a scraper layer and a shifting query mix are the others, and no chart separates them for you.

How to find high-impression, low-click pages in Search Console

Filter for pages that already rank where people look, which is the opposite of sorting by impressions. Open the Performance report, set the date range to the last three months, and work through it in this order.

1. Set Search type to Web first. GSC defaults to Web, but if anyone has changed it, image search and video results are mixed into the totals, and those behave nothing like ordinary Google search results.
2. Open the Queries tab and add the Average position column, then filter to positions under 11. Everything below that is a ranking problem and belongs on a different list.
3. Sort the remaining search terms by impressions, descending. These are the queries where a searcher read your result on page one of the search results page and picked something else.
4. Switch to the Pages tab and repeat, so you can tell a page that underperforms everywhere from one query that underperforms on one page.
5. Check each surviving query against the live result page. You are looking for what sits above you and whether an AI Overview, a featured snippet or an image pack is absorbing the click.
6. Compare the query against your title and first paragraph. If the page answers a different question from the one being asked, no rewrite of the title will fix it honestly.

That position filter is the whole method. Without it you're reading deep-position noise, and the handful of pages worth an hour of work sit buried under several hundred rows that can't respond to anything you do.

When is rewriting a title or meta description worth doing?

Rewrite when the page ranks in roughly the top ten and the current title answers a different question from the query. Those are the conditions where a searcher is reading your result and deciding, which is the only moment a title can change anything.

Don't bother when the page sits past about position 20, when the query has an AI Overview answering it completely, or when the mismatch is between the query and the page rather than the query and the title. That third case is the one that catches people. A sharper title then attracts clicks the page can't satisfy, and the bounce tells Google exactly what you did.

Two mechanics are worth knowing before you start. Google rewrites a large share of title tags anyway and draws on the h1 when it does, so a title and h1 that agree with each other survive more often. Meta descriptions are not a ranking input and Google frequently replaces them, though the ones that survive are usually the queries where your framing beat an auto-extracted snippet.

Set expectations honestly with whoever asked. SearchPilot, which runs SEO split tests on large sites, reports that most experiments come back negative or inconclusive, so a title rewrite is a reasonable bet rather than a repair.

Where the page ranks too deep for a rewrite to help, internal linking is the lever with more evidence behind it. Pointing established pages at the one that is stuck moves position rather than presentation, and position is the binding constraint at that depth.

How to tell automated impressions from human ones

Look at the shape of the query set rather than the volume. Rank trackers, scrapers and monitoring tools leave patterns that human search does not, and four signatures are reliable enough to sort on.

• A combinatorial grid. Queries that read as every permutation of the same few words, with the modifiers swapped. People do not generate perfect grids.
• Search operator syntax. Anything carrying site:, quoted phrases with long -site: exclusion lists, or a competitor's brand name in that form is a monitoring tool, not a buyer.
• Uniform deep positions with no spread. Genuine long-tail keywords vary in position. A query family sitting flat between positions 45 and 70 in the SERP is being polled, not searched.
• Zero clicks across a large sample. Any real query set produces the occasional click. Thousands of impressions and none at all is a strong tell when it coincides with the three above.

You can't fix any of this, and you don't need to. What identifying it buys you is that you stop reporting it as visibility and stop treating its CTR as a failure. Server logs answer the related question of which crawlers actually fetched the page, and tracking AI bot activity covers how to read them.

What impressions with no clicks cannot tell you

Four limits, and they matter because this metric invites confident conclusions it cannot support.

• Search Console will not tell you which surface an impression came from. The generative AI report separates AI features from the total, but within ordinary web results an image pack, a sitelink and a blue link at position 60 are indistinguishable.
• The newest data is preliminary. Google says so in the Performance report itself, and the figures rise as the days finalise, so a dip at the right-hand edge of the graph is usually the graph and not the site.
• A falling CTR is not evidence of decline. It is a ratio, and its denominator has moved for three separate reasons since May 2025.
• Query-level data is sampled. The rows you can see do not sum to the totals, so any decomposition built from them is a large sample and should be described as one.

The one thing this data does establish well is indexing. A page earning impressions is a page Google has stored and is willing to show, which is a different question from whether it ranks, and checking when Google indexed a page covers how to read that properly.

How RankX AI reads Search Console impression data

RankX AI connects to Search Console and splits the Performance report by position band, so the queries that can respond to a change are separated from the ones that cannot before anyone reads a CTR figure. That is the split the position-band table earlier in this article describes.

It also tracks AI Overview citations and brand mentions across assistants separately from Google impressions, because those are different questions with different evidence behind them. Measuring AI search visibility sets out that framework, and query fan-out explains why one prompt turns into many searches in the first place.