Clicks Falling but Impressions Rising: What It Actually Means

If your Search Console chart shows impressions climbing while clicks slide, you are seeing what analysts call the Great Decoupling. For most of search history the two lines moved together: more impressions meant more clicks. Since early 2025 they have pulled apart. The dominant cause is Google AI Overviews and the wider shift to zero-click search, where the user gets the answer on the results page and never clicks through. Your page still logs the impression, but the AI Overview absorbs the click. There are also non-AI causes, from a genuine Google logging error to dead-zone rankings, so the first job is to work out which one is hitting you. For an ad-funded site the stakes are simple: an impression earns nothing, only a click that lands a real reader on your page loads your ads.

Why are my clicks falling while impressions are rising in Search Console?

The short answer is that Google now answers many searches on the results page itself, so your listing is seen without being clicked. To understand why that shows up as two lines pulling apart, you have to look at how Google actually counts the two metrics.

Per Google's own Search Console definitions, an impression is counted when a user has seen, or potentially seen, a link to your site. For an AI Overview, the link only needs to be scrolled or expanded into view to count. A click, by contrast, is counted only when the user is sent to a page outside Google. Reading the AI-generated answer, or ending the session satisfied, produces an impression but no click.

The mechanism that makes this so damaging is positional. The entire AI Overview occupies one SERP position, and every source link inside it is assigned that same position. Unlike a featured snippet, which carries a single link, an AI Overview can cite many pages and hand each one the same strong average position. So your page can register an impression at, say, position 2 or 3 inside an AI Overview while almost never being clicked. Mechanically, that pushes impressions up and clickthrough rate down at the same time. That is the core of the Great Decoupling: the impression and the strong position survive, the click does not.

The behavioural evidence is stark. The Pew Research Center tracked 68,879 unique Google searches from 900 US adults in March 2025. On searches that contained an AI summary, users clicked a traditional result just 8% of the time, versus 15% on searches without one, roughly half the click rate. Only 1% of searches with an AI summary produced a click on a link inside the summary itself. And users ended their browsing session more often when an AI summary appeared: 26% of the time, versus 16% without. That last figure is the clearest picture of zero-click behaviour. The answer was enough, and the journey stopped on Google.

The scale of the CTR hit is quantified elsewhere. Ahrefs' December 2025 study of 300,000 keywords found that the presence of an AI Overview now correlates with a 58% lower clickthrough rate for the top-ranking page, up from the 34.5% reduction the same team measured in April 2025. The damage is worst where you are strongest: minus 58% at position 1, minus 50.8% at position 2, tapering to minus 19.4% at position 10. On keywords carrying an AI Overview, Ahrefs found position-one CTR fell from 0.073 in December 2023 to 0.016 in December 2025, a collapse of nearly four fifths over the two years across which AI Overviews rolled out.

Is falling CTR always caused by AI Overviews?

No. AI Overviews are the headline driver, but treating them as the only explanation will lead you to misdiagnose your own data. Two things complicate the picture.

The first is that AI Overview prevalence itself moves around, so how much decoupling you see is not constant. Semrush's analysis of more than 10 million keywords showed AI Overviews on 6.5% of queries in January 2025, peaking just under 25% in July 2025, then falling back to under 16% by November 2025. That matters for trend analysis. A rise in your impressions can partly reflect Google expanding AI Overviews across more of your queries, and a dip can reflect Google retracting them, rather than any change in real demand or in your content. Month-over-month reads need this context, because the coverage line is not monotonic.

The second is a genuine, non-AI reporting fault. According to SOCi's write-up of the issue, Search Console overstated impression counts from 13 May 2025 onward. Google disclosed the fault around 3 April 2026 and began rolling out a fix, with no historical backfill. That means some of your "impressions rising" across that window is a pure reporting artefact, and it makes year-over-year impression comparisons spanning that period unreliable. If your impression spike sits inside that window, discount it before you draw conclusions.

What else can cause impressions to rise without clicks?

Beyond AI Overviews and the logging error, several benign, everyday causes inflate impressions without adding clicks. None of them is a crisis, but all of them muddy the top-line trend.

• Dead-zone rankings. When you rank on the second or third page, roughly positions 11 to 50, you appear in the results often enough to log impressions but rarely get clicked. A page drifting into the dead zone adds impressions and depresses CTR at the same time, with no AI Overview involved at all.
• Broad or seasonal query rotation. As search demand shifts across the year, your pages start matching more low-intent, tangential queries. Those add impressions that were never likely to convert.
• International or translated SERP inflation. Appearing on translated or international results pages adds impressions from audiences that almost never click through to your page.

The practical read for a publisher is the same in every case: measure net clicks, not impressions or publish count. Impressions rising is only good news if clicks follow. If they do not, the extra visibility is either being absorbed by an AI Overview, sitting in the dead zone, or arriving from low-intent and international queries that will never load an ad.

How do I diagnose this in Search Console?

Diagnosis means splitting the top-line trend into its parts, because the headline impression number blends several effects together. AI-feature traffic is folded into the standard "Web" search type in the Performance report, so AI Overview and AI Mode impressions are already mixed into your headline impression totals. You cannot separate them just by looking at the main chart.

Google's dedicated Generative AI performance report launched in June 2026, per Google's Search Central blog. It is useful but limited. In its first version it surfaces impressions inside AI features while stopping short of full click, position, CTR and query data, so it confirms that you are appearing in AI features without telling you how many clicks those features cost you. That gap is why publishers still cannot directly quantify clicks lost to AI inside Search Console, and why manual segmentation matters.

Work through this checklist:

1. Segment your top queries by AI Overview presence. Identify which of your highest-impression queries now trigger an AI Overview. If rising impressions are concentrated on AI-Overview queries, that points squarely at zero-click absorption rather than lost rankings.
2. Check your average-position band. A flood of low-value impressions in positions 11 to 50 inflates impressions without clicks. If your added impressions cluster in the dead zone, the cause is ranking drift, not AI.
3. Account for the logging error window. If your impression rise falls between 13 May 2025 and the April 2026 fix, treat some of it as a reporting artefact rather than real visibility, and avoid year-over-year impression comparisons across that span.
4. Separate search referrals from AI-engine referrals in GA4. Classic organic sessions and AI-referred sessions behave and monetise differently. Splitting them tells you whether your real, ad-loading traffic is holding up even as the impression line climbs.

The metric that survives all of this is net clicks, and, increasingly, your citation share in AI answers. A worked example makes the point. Suppose a guide gets 100,000 impressions and 8,000 clicks one month, then 130,000 impressions and 6,500 clicks the next. Search Console will show your visibility improving by 30%. Your actual reader traffic, and therefore your ad-loading sessions, fell by almost a fifth. Reading the impression line alone would have told you the opposite of what happened.

Ramprt is built for exactly this split. It reads your existing GA4 with no new tracking script and gives you a clean read of what AI is doing to your traffic: AI referrals by engine, what AI-referred traffic earns versus your site average, and the click decoupling behind the top-line numbers. You can try the free demo to see the AI tab against sample data before you connect your own.

How can I win the clicks back?

There is no switch that turns an AI Overview off, so recovery is about shifting where you compete and how you measure success. A few directions follow directly from the mechanism.

First, change the KPI. Move the scoreboard from impressions and average rankings to net clicks and real ad-loading sessions. Because display and native revenue is earned on page sessions and not on search impressions, an impression inside an AI Overview loads no ads for you. A chart where impressions climb and clicks fall is a leading indicator of falling ad-funded pageviews, even while Search Console "visibility" looks healthier. Tracking net clicks stops the top-line trend from flattering you.

Second, work the position bands that still pay. The Ahrefs breakdown shows the CTR hit is smallest lower down the page, minus 19.4% at position 10 versus minus 58% at position 1. Queries where you sit in the positions 11 to 20 range and can climb into spots that still earn clicks are more valuable than chasing a number-one position that an AI Overview will hollow out. Prioritise the near-miss rankings on commercial and high-intent queries.

Third, prefer queries and formats that resist zero-click. Pew's data shows AI summaries appeared on around 18% of searches in the study, so a large share of search still runs without one. Depth, first-hand data, tools, and content that a reader needs to open the page to use are harder for an AI Overview to fully satisfy. That is where a click still has a reason to happen.

Fourth, accept that citation share is now part of the game. A citation with no click still shapes brand demand, even though it does not load an ad. Tracking how often AI answers are built on your pages tells you whether your content is still the source, which is a leading indicator of demand even when the immediate click is gone. If you want to go deeper on that side of the shift, our companion piece on what AI Overviews are doing to your traffic covers the referral and revenue read in more detail.

The honest position is that some of these clicks are not coming back in their old volume, because the behaviour has changed at the user level. Pew's session-ending figure, 26% versus 16%, is a behaviour shift, not a ranking problem you can out-optimise. The win is to measure the traffic that still monetises, defend the queries that still click, and treat the impression line as a diagnostic input rather than a goal.

What is the revenue read for ad-funded publishers?

It is worth stating the bottom line plainly, because it is easy to be reassured by the wrong chart. For a Raptive, Mediavine or AdSense site, an impression earns nothing. Only a click that lands a real reader on your page loads your ads. So a chart where impressions climb and clicks fall is a direct warning of falling ad-funded sessions, even though your Search Console visibility looks stronger. The correct response is to diagnose by segmentation: AI Overview presence on your top queries, your average-position band, and your referral source, search versus AI engines, rather than trusting the top-line impression trend. Get those three cuts and you will know exactly which version of the decoupling you are looking at, and what, if anything, you can do about it.