AI Overviews France: Who Really Loses 23 % of CTR?
Summarize this article with AI
The -23,1 % Is Not Universal
I look at 15 sites a week. They all share the same problem : they read an aggregate number and think they are safe.
On July 22, 2026, France received AI Overviews. For the first time, we could see before and after without guessing. Ahrefs was already tracking 963 domains in Search Console, 28 days before launch and 9 days after. Juliette Begue and Xibeijia Guan published the data on September 16, 2026.
The number circulating : -23,1 % CTR. Maybe not yours. Across all 963 domains, CTR dropped only 3,4 %. On domains most exposed to AI Overviews, the drop climbs to 23,1 %. Six times more.
See the gap ? An algorithm update does not hit everyone the same way here. Exposure is differentiated.
So who really loses ? The site selling transactional products comes out ahead. The site answering brand queries does too. Those who lose are sites that held the monopoly on informational clicks. Sites whose page was the answer, before Google summarized it.
Segmenting the 963 domains shows the key point : your risk depends on your query profile. I repeat this to every client : before allocating any budget, measure your exposure rate. Not the market average.
I’ll come back to the -0,9 slope in the next section. It is more useful than any global percentage.
The -0,9 Slope: Your Exposure Becomes Your Loss
Ahrefs compared CTR change and exposure rate for each domain. Result : a slope of -0,9. Each point of AI Overview exposure costs about one point of CTR.
If 15 % of your queries trigger an AI Overview, expect to lose about 15 % CTR. It is rough. It is actionable. Pull your exposure rate from Site Explorer and you have your order of magnitude.
Careful : a slope is not a prediction. Some highly exposed sites do fine. Others, barely exposed, collapse for reasons unrelated to AI. Exposure rate is a risk factor, not a verdict.
But the relationship is stable across 963 domains. It does not depend on sector, size, or authority. It depends on the volume of queries where Google decides to answer in your place.
My advice : read this slope as calibration, not condemnation. You know roughly how much you will lose. So you know how much to invest to compensate. Full clarity, even when the number is bad, beats a reassuring aggregate.
I observe among my clients : those who know their exposure rate to within 0,1 make better decisions. The rest react six months late.
A concrete example. An e-commerce site with 1 000 tracked queries. 200 display an AI Overview. Exposure rate : 20 %. Slope -0,9 : expected CTR loss around 18 %. If this site gets 10 000 clicks per month, it would lose 1 800. Not a catastrophe. But a clear signal. You then budget compensation on channels that remain clickable.
Voici la répartition réelle des 963 domaines suivis par Ahrefs. Une minorité encaisse des pertes massives, mais presque un site sur deux perd plus d’un dixième de ses clics en neuf jours.
Qui perd vraiment : la distribution des pertes de clics
48 % des domaines perdent plus de 10 % de leurs clics, 5 % en perdent la moitié
The Brutal Distribution: 1 in 20 Sites Loses Half Its Clicks
The global median : -5,7 % CTR and -8,8 % clicks. Reassuring ? No. The distribution is brutal.
Daily, 48 % of domains lost more than 10 % of their clicks. 19 % lost more than 30 %. 5 % lost more than half. Almost one site in two lost more than a tenth of its clicks in nine days. One in twenty lost half.
| Daily click loss | Panel share |
|---|---|
| More than 10 % | 48 % |
| More than 30 % | 19 % |
| More than 50 % | 5 % |
Not just a few unlucky sites dragging down the average. 82,2 % of highly exposed domains lost CTR. At that exposure level, coming out ahead is the exception.
The hardest-hit sector : health. Sites answering » symptoms « , » treatment « , » difference between X and Y » see their clicks melt away. Google summarizes, the user no longer clicks. It is mechanical.
For e-commerce, the reading is different. If you sell products, your category pages and product pages are less exposed to AI Overviews than your informational blog posts. The risk concentrates on your supporting content, not your transactional funnel. But that content often brought you status and brand awareness. We come back to that with dopamine.
In short : do not take comfort in the median. Look at your own click distribution by query type. The devil is in the segments, not the average.
Impressions Up, Clicks Down: The Decoupling
The most extreme case in the French panel : +183 % impressions, -1,9 % clicks. CTR is divided by more than three.
Read carefully : pages still appear, sometimes more often. They are simply clicked less. The AI Overview does not remove you from the SERP. It makes you redundant within it.
Ahrefs found the same pattern in its global study in late 2025. It is the » great decoupling « . Visibility no longer converts into clicks. Impressions no longer deliver anything.
For an e-commerce manager, the consequence is direct : your impressions report is worthless as a success indicator. Track CTR by query, by page, by intent. A page that gains impressions and loses clicks is not growing. It is spinning its wheels.
I often say in audits : ranking is no longer the end goal. The citation in the generated answer is the new first position. If Google does not cite you, your organic #1 position only confirms the AI.
The worst part ? This decoupling will intensify. Today, the median exposure rate is 5,4 %, and 29 % of domains are already above 10 %. Across all markets since 2024, these numbers have only increased. France will be no exception.
Look at your data. If impressions climb but clicks stagnate or fall, you are in this decoupling. No need to wait for Ahrefs’ follow-up study in a few months. You already see it in your Search Console.
Dopamine Down: Losing Status, Not Just Clicks
The DOSE framework I use with Guillaume Attias (BMO Academy) explains what is at stake : you lose clicks, and also status.
Serotonin, in the DOSE framework, is authority status. The site that was the reference on an informational question held a monopoly. The user typed » best moisturizer for dry skin » and clicked on your guide. You were the answer.
With AI Overviews, Google becomes the answer. Your reference status evaporates. Your content has nothing to do with it : the interface itself changed.
The most exposed sites are those that held this monopoly on informational clicks. Their CTR drop is also a status drop. More content does not repair it. What repairs it : citations in the ecosystem Google uses to generate its answers.
For an e-commerce manager, this is a shift : if you built your authority on a blog that answered broad questions, you lose that trust-acquisition channel. Your product pages are not directly threatened, but the brand awareness that fed them is shrinking.
I observe among my clients that those with a strong blog on informational queries see their CTR melt faster than their category pages. Shift your effort toward citations, the sources AI Overviews pick up, the entities that feed the knowledge graph.
And this status is measurable. In citations, not rankings. New playing field.
Avant d’allouer le moindre budget, mesurez. Voici le chemin concret pour transformer votre taux d’exposition en ordre de grandeur exploitable.
Estimer votre propre exposition en 4 étapes
Du taux d’exposition agrégé à votre risque réel de CTR
What You Gain by Measuring Before You Allocate
Stay calm and pull your exposure rate, query by query.
Go to Ahrefs Site Explorer, filter by » AI Overview » in SERP features. Separate your transactional queries ( » buy « , » price « , » product review « ) from informational queries ( » how « , » why « , » difference « ).
Apply the -0,9 slope to your informational exposure rate. If 20 % of these queries trigger an AI Overview, expect to lose about 20 % CTR on that segment. It is an order of magnitude, not a fine prediction. It at least tells you whether you are in the 82,2 % that lose or the 17,8 % that come out ahead.
Then allocate. If your exposure is low, continue your classic SEO. If it is high on informational content, invest in citation presence : be cited by the sources Google picks up, structure your data, appear in third-party databases, studies, and comparison sites.
Full clarity : you know if you belong to the most exposed segment before spending a single euro on SEO or GEO.
The -23,1 % hides very different profiles. A marker, nothing more. And you can measure it today.
What percentage of your queries trigger an AI Overview right now?
Get Your AI Overviews Exposure in 30 Minutes
In the audit, I segment your queries, calculate your exposure slope, and identify the pages that will lose the most. You leave with a citation plan, not guesswork.
Book a strategic call — 45 minFrequently Asked Questions
What exactly is the -23,1 % CTR drop?
Ahrefs measured this drop on the French domains most exposed to AI Overviews. The reading dates from the 9th day after launch, July 22, 2026. If we widen to the 963 domains in the study, the average drop is limited to -3,4 %.
How do I know if my site is exposed to AI Overviews?
In Ahrefs Site Explorer, I filter queries by the « AI Overview » SERP feature. The percentage of queries that trigger one is my exposure rate. Then apply the -0,9 slope to estimate CTR loss.
Does the -0,9 slope apply to all sectors?
Yes. Ahrefs calculated it across 963 domains, all sectors combined. The relationship is stable. It remains a risk factor. Not an exact prediction. And some highly exposed sites come out better than others.
Are e-commerce sites more exposed than editorial sites?
It depends on the query type. On product pages and catégories, I see fewer AI Overviews than on guides and articles. The risk concentrates on your blog and advice pages.
What should I do if my site is in the most exposed segment?
Get cited by the sources Google picks up in its AI Overviews. Structure your data. Appear in third-party databases and studies. Authority is earned through citations. Ranking alone is no longer enough.
