AI Overviews cut French site clicks by 23% within days
The French CTR drop matches every prior AI Overviews rollout almost exactly, which means citation, not click-through, is now the metric that counts.
Key takeaways
- Ahrefs measured a 23.1% CTR drop for the most exposed domains nine days after France's AI Overviews launch on 22 July 2026.
- The figure matches prior rollouts in the US, UK and Germany, disproving the theory that CTR loss was a temporary novelty effect.
- Informational and authority-seeking content, the kind published by banks, UN agencies and standards bodies, sits disproportionately in the most-exposed cohort.
- CTR is no longer a reliable health metric; citation inside the AI Overview itself is the KPI that now matters.
- Financial services, multilateral and industrial content teams still reporting CTR to leadership are tracking the wrong number.
France got AI Overviews on 22 July 2026. Nine days later, per Ahrefs, the 963 domains most exposed to the feature in Google Search Console had lost 23.1% of their click-through rate, measured against the 28 days prior. Nine days. Not a quarter, not a slow bleed over a fiscal year. A market woke up one Wednesday with a search feature it hadn't asked for and by the following Friday had a fifth of its organic clicks gone.
The interesting part is not that this happened. It is that it happened exactly as it did everywhere else, and everyone in search marketing still treats each rollout as news.
A pattern with no exceptions
AI Overviews landed in the US in May 2024. Similar CTR damage followed, documented by multiple independent trackers over the following year: figures in the 15 to 25% range depending on query category, worse for informational intent, less severe for transactional. The UK got it in late 2024. Same story. Germany, same story. Now France, and the number lands squarely inside the range everyone should have expected: 23.1%.
This is worth dwelling on because it kills a comforting theory that circulated after the US launch: that the CTR collapse was a peculiarity of American search behaviour, or a temporary novelty effect that would fade as users learned to scroll past the summary. Eighteen months and five markets later, the novelty-effect theory is dead. AI Overviews does not erode clicks because users are confused by something new. It erodes clicks because it answers the question. That is a structural feature of the product, not a transitional glitch, and it is not going to unwind itself as French users acclimatise. Ahrefs' own data across earlier rollouts shows the effect holding steady months after launch, not recovering.
For brands that built visibility strategy on the assumption that "things will normalise once people get used to AI Overviews," France is the fourth or fifth data point refuting that assumption in a row. At what point does a brand's SEO plan stop pricing in reversion to the mean that isn't coming?
Why the exposed 963 matter more than the average
The headline number understates the story for anyone selling complexity: financial services explainers, multilateral policy briefings, ISO-style standards guidance, industrial technical documentation. These are exactly the query types Google's AI Overviews target first, because they are exactly the query types that reward a synthesized answer over a list of blue links. A user asking "what does Basel III liquidity coverage ratio require" or "how does the Sendai Framework define disaster risk reduction" does not want ten tabs. Google now gives them a paragraph, stitched from sources the user never has to click through to see.
That means the domains publishing this kind of content, banks' research arms, UN agency knowledge hubs, standards bodies, industrial group whitepapers, sit disproportionately inside Ahrefs' "most exposed" cohort, not the "average across 963 domains" figure. The 23.1% is a floor for informational, authority-seeking content, not a ceiling.