AI Overviews France: who really loses CTR? By exposure
Summarize this article with AI
July 22, 2026: a measurable shock, not an impression
July 22, 2026. 963 domains tracked. -23.1% click-through rate for the most exposed. The numbers are there, measured. They come from Ahrefs, published by Juliette Begue on September 16, 2026.
France switched to AI Overviews that day. For once, no one had to guess the effect. The rollout was clean. 28 days before, 9 days after. Google Search Console data already existed.
I look at Search Consoles every day. What I see with my clients looks like what the study shows. Clicks are dropping. Positions are holding. A silent erosion.
The overall number is -3.4% CTR in nine days. That’s almost nothing, said like that. But that number is an average that kills the diagnosis. The most exposed domains lost 23.1%. Six times more. Segmentation changes everything.
We can no longer reason by overall traffic. We reason by exposure level. That’s the only way to isolate who really loses CTR.
Result: 82.2% of highly exposed domains lost CTR. We’re far from a few unlucky sites. At that exposure level, coming out unscathed becomes the exception.
SEO is managed at the query level, not the domain level. A site with 100,000 sessions may have 5% of queries exposed. Another with 10,000 sessions may have 40%. The two don’t live the same reality.
The Ahrefs study shows it. The median exposure rate is 5.4%. But 29% of domains are already above 10%. The extremes weigh heavily.
I sell measurement. The kind that turns a scare into an action plan.
My observation is simple. What matters is the exposure level: how exposed am I?
The 28 days / 9 days method: why it changes everything
Google Search Console tells the truth. But it mixes everything. Core updates, seasonality, tracking changes. To isolate the effect of AI Overviews, I needed a clean window.
Ahrefs chose 28 days before July 22, 2026. Nine days after. 963 French domains. Sites of all sizes, from all sectors.
Why 28 days before? Because it smooths daily variations. Weekdays. Weekends. Seasonal peaks. Nine days after are enough to see the shift. The rollout cut clean.
The number is clear. The slope between AI Overviews exposure and CTR loss is -0.9. Each additional exposure point costs about one point of click-through rate. The algorithm has nothing to do with it. The interface redefines the click.
With my clients, I tested this logic. An e-commerce site with 800 pages measured 14% of exposed queries. CTR dropped by 11% over thirty days. The order of magnitude holds.
This 28d/9d method has become my diagnostic tool. I no longer rely on feeling. I measure exposure. Then I segment. That’s where the sorting happens.
You don’t need 963 domains. Your Search Console is enough. Export the 28 days before July 22. Export the 9 days after. Compare CTR by page. You’ll see the same shift, at your scale.
Full clarity. The numbers speak.
Segment by exposure, not by feeling
Segmentation is the heart of the matter. The exposure rate is calculated simply. How many of your queries trigger an AI Overview? Divided by the total number of queries. That gives a percentage.
At Ahrefs, the median rate is 5.4%. But the distribution is brutal. The most exposed don’t play in the same league.
| Exposure level | Average CTR loss | Share of affected sites |
|---|---|---|
| Low (under 5%) | Almost stable | Silent majority |
| Medium (5 to 15%) | Around -5 to -10% | A growing share |
| High (above 15%) | -23.1% in nine days | Minority but massive impact |
This table is for prioritizing. Nothing else.
82.2% of highly exposed domains lost CTR. The risk is real, but not universal. Some get through. Others take the hit.
I distrust averages. Across the panel, the median is -5.7% CTR and -8.8% clicks. That reassures at first glance. Except that this median hides a violent distribution.
Nearly one in two sites lost more than 10% of their daily clicks. One in twenty lost half. In nine days. Not in a quarter.
Hence my segmentation. You need altitude to fly a plane. You need an exposure rate to run a site.
Three site profiles facing AI Overviews
In the field, I see three profiles. The first has almost nothing. The second feels a tremor. The third takes the hit.
Profile 1: the spared. Exposure rate under 5%. Its queries are transactional, navigational, local. AI Overviews don’t appear for « plumber Lyon 3 » or « customer login ». The CTR loss is minimal, often under 2%. This is the profile that sleeps well.
Profile 2: the rubbed. Exposure rate between 5 and 15%. Informational and comparative queries. Example: « best robot vacuum 2026 ». The AI Overview summarizes the answer, the click erodes by 5 to 10%. The site remains visible, but CTR slowly melts.
Profile 3: the decoupled. Exposure rate above 15%. Often media, health sites, encyclopedias. Pages still rank, sometimes better. Impressions rise. Clicks fall. That’s the great decoupling.
The Ahrefs study confirms the worst sector: health. Medical queries massively trigger AI Overviews. Publishers in this vertical lose more than others. Logical. Google answers directly, the page becomes redundant.
With my clients, I see a fourth variable. Content type. Pure information pages suffer more than product pages. Pages that sell resist better. Provided they have clear positioning.
Segment your site according to these three profiles. Group pages by theme, by intent, by exposure rate. You’ll see your pockets of risk appear. That’s where you need to act.
Exposure is a risk factor, not a verdict.
Une image vaut mieux qu’un long discours sur le découplage. Sur le domaine le plus exposé du panel, les impressions décollent pendant que les clics reculent.
Le découplage en chiffres
Impressions en hausse, clics en baisse : le cas extrême du panel Ahrefs.
Why impressions rise and clicks fall
Decoupling is the most destabilizing thing. Rankings don’t move. Impressions rise. Clicks collapse.
In the French panel, the extreme case shows +183% impressions for -1.9% clicks. CTR is divided by more than three. The page is still there. Except it’s now useless.
The AI Overview keeps your page in the SERP. It makes it useless within the same SERP. The user has their answer before clicking. Why would they click?
Content has nothing to do with it. The problem is in the search results page architecture. Google added a layer. It captures the click.
I insist: impressions say nothing about a site’s health. You can show more impressions and lose revenue. CTR is the only indicator that matters.
The overall data speaks for itself. The median is -5.7% CTR. Daily clicks lost: 48% of sites lost more than 10%. 19% more than 30%. 5% more than 50%. In nine days.
Believing that « SEO is dead » is a mistake. Believing that nothing moves is also a mistake. We need to rebuild.
Guillaume Attias’s DOSE framework taught me that. Separate the observable from the decision-making. Observe without panicking. Decide without waiting.
The impressions/clicks decoupling is an observable. The decision is knowing which pages to fortify.
Avant d’écrire un mot de plus, mesurez. Voici la méthode que j’applique chez mes clients, en six étapes simples.
Mesurer son exposition en 6 étapes
Le workflow pour savoir si vous êtes dans le profil découplé.
What you gain by measuring your exposure now
Before writing more content, measure your exposure rate. Without this number, you’re navigating in fog.
I apply this method with my clients. Export your Google Search Console queries for 28 days before July 22. Then the 9 days after. For each query, identify if an AI Overview appears. Ahrefs Site Explorer filters by SERP features. It’s the fastest.
Calculate your exposure rate. Queries with AI Overview divided by total queries. Compare with the panel: median 5.4%. Above 15%, you’re in the decoupled profile. You benefit from acting fast.
Then, segment by page. Which pages carry the exposed queries? These are your pages on borrowed time. Often informational pages, guides, definitions. They don’t deserve the same strategy as your transactional pages.
Countermeasures exist. I won’t detail them here. The priority order is clear. First measure. Then segment. Then decide.
I’ve seen sites react within fifteen days. Reduce their dependence on purely informational queries. Strengthen the pages that convert. Test formats that capture the click even when the AI Overview answers. CTR rises again.
Exposure will continue to increase. Today 29% of domains are above 10%. In six months, it will be more. Waiting is choosing to suffer.
You don’t manage a site by feeling. You manage it with numbers. Exposure-based segmentation is your first number.
Is your site in the most exposed quartile?
AI Overviews exposure audit
I measure your exposure rate on your first 500 queries and identify the pages on borrowed time. You leave with an action plan.
Book a strategic call — 45 minFrequently Asked Questions
How do I calculate my AI Overviews exposure rate?
Export your GSC queries. Identify those that trigger an AI Overview in the SERP. Divide that number by the total number of queries. In Ahrefs Site Explorer, the SERP features filter saves you time.
Does a 23.1% CTR loss affect everyone?
No. It’s the average for the most exposed domains. The overall median is -5.7%. It all depends on your exposure rate.
Do AI Overviews remove positions?
No. Impressions hold, sometimes they even rise. What disappears is the click. Pages remain ranked, but they duplicate in the SERP.
Which sector loses the most with AI Overviews in France?
Health. Medical queries massively trigger AI Overviews, which strongly reduces CTR for publishers in this vertical.
Is the -0.9 slope a reliable prediction?
A risk, not a sentence. If 15% of your queries are exposed, count on about 15% less CTR. I’ve seen highly exposed sites come out without damage.
