AI Overviews France: What the 963 Domains Tracked in GSC Reveal
Summarize this article with AI
A client calls in late July. He read the study. He wants his number.
On July 22, 2026, AI Overviews arrive in France.
The following Monday, a client calls. He runs a health site. He read the Ahrefs study published on September 16, 2026. He asks: « And for my site, what does it show? »
I have a method, not an immediate answer.
First, the framework. 963 domains tracked in Search Console. 28 days before rollout. 9 days after. A first reading of the structure, going beyond the global figure.
According to the study led by Juliette Begue, France Marketing Manager at Ahrefs, and Xibeijia Guan, data scientist, the domains most exposed to AI Overviews lost 23,1 % of click-through rate in nine days.
23,1 % fewer clicks, measured over nine days.
I look at that number like an engineer. It doesn’t say everything. It says where to start.
Why now? For once, France has a precise launch date. In most markets, AI Overviews arrived at the same time as algorithm updates and seasonal variations. We spent months debating. Here, we measure before and after: 28 days before, 9 days after.
Clarity is total.
The panel is clean: 963 domains. .fr sites and international players positioning themselves in France. 29 % already exceed 10 % AI Overviews exposure rate. That’s concrete.
In other markets, the arrival of AI Overviews mixed with helpful content updates, core updates, seasonal adjustments. Isolating the effect was impossible. France provides a clean laboratory: a before, an after, domains already tracked.
The Ahrefs study goes beyond the average. It traces a spectrum. And that spectrum is what matters to an SEO manager.
The health client resembles many others. In the weeks following July 22, fifteen sites are reviewed each week. The same pattern appears. Stable impressions, clicks slipping.
All of this can be measured. And it’s unsettling.
Derrière la médiane rassurante se cache une distribution brutale. Voici comment se répartissent réellement les 963 domaines suivis dans Search Console.
La perte de clics n’est pas une moyenne
Répartition des 963 domaines selon la perte de clics quotidiens en 9 jours
23,1 %. Not an average. A spectrum.
First reading: the average reassures wrongly.
Across the whole panel, the median click-through rate drop is 5,7 %. The median click drop is 8,8 %.
But behind that median, the distribution is brutal.
| Daily click loss | Share of panel |
|---|---|
| More than 10 % | 48 % |
| More than 30 % | 19 % |
| More than 50 % | 5 % |
In nine days.
82,2 % of highly exposed domains saw their click-through rate drop. At that exposure level, coming out undamaged is the exception.
The numbers are unambiguous. Almost one site in two lost more than 10 % of daily clicks. One in twenty lost more than half. In just nine days.
The study says the opposite of a uniform impact: the panel splits into two camps. Those who fall behind. Those who hold up.
The most affected sector: health.
Logical. A health query is often informational. A three-line summary is enough to answer. The click becomes a luxury.
The .fr domains fall behind more than international players. The median click loss for .fr is 12,3 %, compared to 8,8 % for the whole panel.
Here, the change comes from a new component of the results page, not from the algorithm.
The panel’s median exposure rate is 5,4 %. But 29 % of domains are already above 10 %. And these numbers have only increased since 2024 across all markets.
Health is the most visible, but it is not the only one affected. Definition sites, educational content, basic recipes follow the same logic.
Another thing: the click drop does not stem from lost positions. The pages remain. They still appear. They are simply clicked less. This is the great decoupling.
The remaining question is where my domain sits on this spectrum.
The mechanism: the page remains, the click disappears
The most extreme case in the study: +183 % impressions for −1,9 % clicks.
In other words: the page is more visible than ever. It generates fewer clicks. Its click-through rate is divided by more than three.
The page keeps its position. It still appears, sometimes more often. Except people click on it less.
The same pattern emerges from the worldwide study conducted at the end of 2025. And it’s the same shape as the « great decoupling »: AI Overview does not remove your page from the SERP. It makes your page redundant inside it.
The relationship between exposure and CTR loss holds. Across the 963 domains, the slope is −0.9.
Slope of −0.9: each additional point of exposure costs about one point of click-through rate.
Plainly: one more point of AI Overviews exposure means roughly one point less CTR.
If 15 % of your queries trigger an AI Overview, expect about 15 % less CTR.
It’s a rough estimate. Some highly exposed sites hold up well. Others, barely exposed, collapse for different reasons. The exposure rate gives a risk level. It decides nothing.
This regularity is useful. Pull your exposure rate: you have your order of magnitude within minutes.
Why this regularity? Because the mechanism is automatic. The more a query displays an AI Overview, the more visible the answer is in the SERP. The user has less need to click. The blue page is still there. It waits. It receives fewer visits.
The study also notes that almost nobody lost positions in late July. Pages continue to appear, sometimes more often. They are simply clicked less.
The case of +183 % impressions for −1,9 % clicks remains the most extreme in the French panel. It stands out. The mechanism is pushed to its maximum there.
We stopped asking « is the AI Overview stealing my traffic? ». We measure the exposure rate. We quantify the loss. We act.
The formula is simple. The diagnosis too. The action, though, requires discipline.
Who falls behind? Who holds up?
The Ahrefs study does not cover every sector. It names one: health.
In my weekly monitoring, this holds up. The sites that fall behind fastest answer simple, factual questions that AI Overview summarizes in three lines.
A health blog explaining the symptoms of a deficiency. A definitions site. A basic recipe site.
At the opposite end, domains that hold up often share one point: they offer information that AI Overview cannot condense without losing information.
- Detailed comparisons with proprietary data.
- Real-life feedback, with photos, measurements, tests.
- Interactive tools, simulators, calculators.
- Transactional content where the user wants to compare prices, stock, delivery times.
A .fr is more exposed than an international player in the same market. The median click loss for .fr is 12,3 %, against 8,8 % for the whole. One possible explanation: .fr domains often cover local and informational queries, less defended by a strong brand.
Holding up comes down to one thing: your position on the exposure spectrum. Size doesn’t enter the picture.
Your domain holds a precise place on that spectrum.
The remaining question: where.
Two typical examples. A natural health site with 400 in-depth articles, strong internal linking, but factual answers. It will be exposed on queries like « magnesium deficiency symptoms ». AI Overview answers in three lines. The click goes.
Beside it, an e-commerce site selling supplements, with detailed product pages, customer reviews, batch photos, price-per-dose comparisons. The user has to click to compare. Exposure is lower, and so is impact.
The sector matters little. The type of answer decides.
The domains that hold up are not smarter than the others. They simply occupy queries where the decision requires a full page, not a summary.
The exposure spectrum serves as a compass. Site size, age, budget change nothing. What counts is the share of queries that trigger an AI Overview.
Guillaume Attias’s DOSE framework: cortisol is not an enemy
Faced with these numbers, the natural reaction is a spike in cortisol. Risk anticipation. Uncertainty. The urge to change everything at once.
I teach DOSE at BMO Academy, Guillaume Attias’s framework. Cortisol is not an enemy. It’s a signal of anticipation. We channel it into action.
In the AI Overviews context, it looks like this.
D: Dopamine. Understand your real exposure. Pull your exposure rate, query by query. A clear map in front of you.
O: Oxytocin. Trust in a framework. The Ahrefs study gives a clear slope. We measure, we act.
S: Serotonin. The pride of having prioritized the right pages. We identify the 20 most-clicked queries that trigger an AI Overview. We protect those first.
E: Endorphin. Move into action quickly, instead of waiting for the next quarterly report.
Nothing mystical here. A decision-making mechanism.
The first action is simple: pull your exposure rate. If you don’t know how, an audit shows it in one session.
Cortisol turns into a plan. The plan executes.
Clarity of numbers reduces cortisol. A number like 23,1 % scares people. A number like « your domain is 7 % exposed across 20 queries » becomes a lever. We know what to do. We know where to start.
DOSE rests on a biological sequence that good decision-makers use without knowing it. Anticipation becomes fuel there.
In an audit, the most productive moment is often when the client becomes aware of their exact rate. Not before. Not after. At that moment, the decision becomes obvious.
Mild cortisol is useful. It pushes you to act. The trap would be to stay in panic and break everything.
We leave everything intact. We measure. We reinforce. We protect.
Inutile de tout changer d’un coup. Voici la séquence que je déroulerais dans l’ordre si j’étais à votre place.
La séquence à lancer dès aujourd’hui
Trois étapes pour transformer le diagnostic en actions
What I would do in your place
The sequence I would run, I would start today, not in six months.
1. Measure. Pull your exposure rate over the last 28 days. Cross your queries with AI Overviews. If 5 % of your queries trigger one, expect an impact of roughly 5 CTR points. If your tool lacks feature segmentation, compare clicks and impressions before/after for the queries where you suspect an AI Overview. The gap gives you an order of magnitude.
2. Isolate the 20 queries that matter. The 20 queries that generate the most conversions, contacts, sales. How many trigger an AI Overview? Look at those first.
3. Create the difference. For each exposed query, add what AI Overview cannot copy: a quantified comparison, a dated field report, proprietary data, a verified review, an interactive tool. A photo of a real-world test. A measurement taken from a specific batch. A price comparison with a date. A review with a purchase date.
4. Strengthen transactional pages. In the study, the impact is stronger on informational content. Pages where the user has to make a purchase or contact decision hold up better. You gain by strengthening their internal linking, by linking to them from exposed content.
5. Track weekly. Visibility is no longer enough. Click-through rate becomes the health indicator. A page that gains impressions and loses clicks is unhealthy.
This sequence takes a few days. It requires precision.
A number beats a debate.
France has a launch date. You have a reference point.
Now you situate your domain on the spectrum.
And you, how many of your queries already trigger an AI Overview?
AI Overviews Exposure Audit
I measure your real exposure rate over your last 28 days, query by query. You leave with the 20 actions that protect your clicks.
Book a strategic call — 45 minFrequently Asked Questions
What exactly did the Ahrefs study measure?
I tracked 963 domains in Search Console: 28 days before the AI Overviews launch in France on July 22, 2026, and 9 days after. Then I crossed the exposure rate with CTR and click evolution.
What does a −0,9 slope mean?
One more point of AI Overviews exposure means one point less click-through rate. I observe the same mechanics: if 15 % of your queries trigger an AI Overview, expect 15 % lower CTR.
Is my French site more exposed than an international site?
Yes. .fr domains show a median click loss of 12,3 %. The full panel is at 8,8 %. The gap makes sense: French players mostly work on local and informational queries, where strong brands are less present.
Do AI Overviews remove positions?
Positions remain. Pages still appear, sometimes even more often. People simply click less. Visibility rises, clicks fall.
What should I do first after this diagnosis?
Look at your exposure rate over the last 28 days. Identify the 20 queries that convert best and trigger an AI Overview. On those pages, add what an AI Overview summarizes poorly: your own data, comparisons with their dates, field feedback.
