Google Rolls Out AI Overviews and AI Mode in France: What Changes for Business Visibility

On 22 July 2026, Google switched on AI Overviews and AI Mode in France. On a growing share of queries, users now get an AI-written answer at the top of the page, before the first blue link. France was the last major market to flip: elsewhere, the experience has been running for two years and the impact data is already in. It does not make comfortable reading for anyone who depends on organic traffic.

What Google actually shipped

Two distinct components need to be told apart, and press coverage routinely conflates them. AI Overviews are an automatically generated block at the top of the standard results page: a few lines of synthesis assembled from several web sources, with links to the pages that were used. AI Mode is a tab in its own right — a conversational search experience where you talk to the engine by text, by voice or from an image, with the ability to chain follow-up questions without starting a fresh query.

Both features run on a version of Gemini tuned for search. AI Mode had already been available in more than 200 countries since late May 2026; its arrival in France completes the rollout in one of the most closely watched European markets from a regulatory standpoint, particularly on the question of press publishers' neighbouring rights. French tech outlet Le Blog du Modérateur lays out the rollout timeline, and Google documents how these features work on its Search product blog.

In practical terms, the results page French users had known for twenty years changed character overnight. This is not a discreet A/B test: it is a rewriting of the implicit contract between Google and the sites that feed it.

The impact numbers, market by market

The upside — such as it is — of going last is that you know what to expect. Several independent studies have measured the effect of AI Overviews on click behaviour, and their conclusions converge.

Ahrefs analysed 300,000 keywords and recorded a 34.5% drop in click-through rate on the first organic position when an AI Overview is present. A refresh of that study, published in February 2026, puts the figure at 58%: the effect has not plateaued, it has worsened in eight months. The share of queries ending without any click at all climbs to roughly 83% when an overview is shown, versus 60% when it is not.

On the behavioural side, the Pew Research Center tracked the actual browsing of 900 US adults across nearly 69,000 Google queries. The verdict: users click a link in 8% of cases when an AI summary is displayed, against 15% otherwise. More telling still, only 1% click the sources cited inside the block. In other words, being cited by the AI does not mean receiving traffic.

News publishers in markets that already have the feature report Google traffic losses of between 20% and 40%, and the June 2026 Digital News Report from the Reuters Institute projects a 43% decline in search-driven traffic for news media over three years. In France, on the very day of activation, a major regional press group told AFP it had lost 30% of its Google traffic. Search Engine Land catalogues further data sets pointing the same way.

Who is really exposed — and who is not

It would be dishonest to declare the death of search marketing. The damage is real but it is distributed very unevenly, and that nuance changes everything for an SME. AI Overviews fire heavily on informational queries — "how does it work", "what is", "what's the difference between". That is precisely the kind of content the AI can summarise without the user needing to go any further: the answer fits in three sentences, it has been handed over, the click no longer has a purpose.

Conversely, transactional and local queries hold up markedly better. When somebody is looking for a provider, comparing prices or wants a quote, a summary will not do: they need to see who is behind the offer, check references, get in touch. An AI block does not replace a commercial relationship. A business whose acquisition rests on service pages, pricing pages and a strong local presence absorbs the shock far better than a publisher that lived off generic explainer content.

For an SME, the real danger lies elsewhere: in company blogs built as informational content factories. The "10 tips for..." articles that used to bring in broad, poorly qualified traffic will mechanically drain away. That traffic was often a vanity metric — its disappearance hurts the dashboards more than the revenue line.

What to do now, concretely

The first useful reaction is to separate informational queries from commercial ones in Search Console, then watch the two curves independently over the next three months. An overall 25% drop in clicks means nothing until you have checked whether the queries that generate inbound enquiries are affected. In most of the cases I observe, they are hit far less than the average. You also have to accept that the impressions/clicks pairing changes meaning: impressions will keep rising while clicks fall, and that is no longer a sign of poor optimisation.

Then comes the substantive work: becoming citable by an AI rather than merely well ranked. The models that generate these answers favour pages whose information is structured, dated, quantified and attributed to an identifiable author. A precise data point with its source, a direct answer placed at the start of a section, correct Schema.org markup (Organization, LocalBusiness, FAQPage, Person) and content that is technically accessible to crawlers now count for more than keyword density. An llms.txt file at the root, still optional, is starting to play the role robots.txt played in its day.

Finally — and this is the most strategic point — it has become unreasonable to rest your acquisition on a single tap whose flow rate a third party controls. Google has just demonstrated that it can redistribute the value of organic traffic through a simple product update. Owned channels (email list, Google Business Profile listing, professional network, referral) and the operational tools that convert visitors already on site into customers regain priority over the race for traffic volume.

A structural shift, not an incident

What played out on 22 July goes beyond search rankings. The web was built on a tacit exchange: sites produce content, engines send visitors. Generative answers short-circuit that reciprocity — the content is consumed, the visitor never arrives. It is the same underlying movement that is pushing companies to embed AI directly into their own products rather than absorb it from the outside, a subject we cover in our overview of web development trends for 2026.

For small and mid-sized businesses, the operational conclusion is a fairly healthy one: less generic content, more evidence, more tools that genuinely serve customers. If you want to know where your visibility is exposed and which AI building blocks are worth the investment in your specific case, our free AI audit delivers the diagnosis in an hour, and our AI strategy consulting takes over once the priorities are clear.