Key takeaways
- AI Overviews split success into two goals: being cited in the answer and earning the click that remains.
- Structure content for extraction—clear early answers, headings, and lists make you citable.
- Add AI-answer visibility as a KPI alongside traffic, using Search Console, third-party tools, and manual tests.
- Opting out of AI Overviews usually means surrendering discovery; presence plus distinctiveness wins.
- Shift from click-led to brand-led content: original research and expertise build authority even without the visit.
With AI Overviews now live in France since 22 July, the ground has shifted for anyone who depends on organic search. These AI-generated summaries sit at the very top of the results page, answering the query before a single blue link gets a click. The uncomfortable reality is that the funnel marketers spent a decade optimising is being quietly rebuilt around them, and content teams need a new playbook.
At Evocative, we've watched this play out across markets that got AI Overviews earlier. The pattern is consistent: discovery increasingly happens inside AI answers, informational traffic falls, and the content that survives is the content the models choose to cite. This isn't a reason to panic. It's a reason to change how you produce, structure, and measure content.
What AI Overviews actually change
The most cited effect is the so-called "funnel collapse." Users get their answer directly in the overview—or increasingly inside ChatGPT and other assistants—without ever visiting a website. Studies from the UK, where AI Overviews landed roughly two years before France, document sharp drops in click-through for publishers, especially on informational and "how-to" queries.
The strategic takeaway is subtle. Rankings still exist, but ranking first no longer guarantees the click. Your goal splits into two: being cited inside the overview (visibility) and earning the click that remains (traffic). Content that does both wins; content optimised only for old-school rankings loses on both fronts.
Rethinking content production for the answer layer
Google's own guidance is refreshingly unglamorous: there is no secret "AI Overviews markup." The same fundamentals that make content genuinely helpful make it citable. But the emphasis shifts in a few practical ways.
- Answer the question cleanly, early. Models extract concise, self-contained statements. A crisp definition or direct answer in the first paragraph is more likely to be surfaced than one buried under context.
- Structure for extraction. Clear headings, short paragraphs, lists, and tables let AI systems lift and attribute specific passages. Ambiguous, meandering prose is harder to cite.
- Demonstrate real expertise. First-hand data, original analysis, named authors, and genuine experience are increasingly the differentiators. Generic content is precisely what the model can already generate itself.
- Go where the model can't. Proprietary benchmarks, case studies, opinion, and depth create reasons for both the AI to cite you and the human to click through.
Measuring visibility in AI Overviews
You cannot manage what you can't see, and traditional rank tracking no longer tells the full story. Build a measurement stack in layers:
- Search Console remains your baseline for impressions and clicks, and it lets you spot the divergence when impressions hold but clicks fall—a tell-tale AI Overview signature.
- Third-party tools are rapidly adding AI Overview and AI Mode tracking, letting you monitor citation share for your priority queries.
- Manual testing still matters. Regularly run your key queries yourself to see whether—and how—you're being quoted.
Track a new KPI alongside traffic: presence and prominence in AI answers. It's the leading indicator of where discovery is heading.
Should you exclude your content?
Google's Search Console offers an option to keep your content out of AI Overviews. For most businesses, opting out is the wrong instinct—you'd surrender visibility in the exact surface where discovery is concentrating. Exclusion only makes sense for narrow cases where clicks are the sole monetisation model and citation delivers zero value. For everyone else, the smarter move is to be present, cited, and distinctive.
The Evocative angle: build brand-led, not click-led
The deeper lesson is that content strategy has to detach from pure traffic dependency. When the click is no longer guaranteed, brand becomes the asset that survives. Being repeatedly cited as a trusted source in AI answers builds authority even when the visit doesn't happen—and that authority converts later, through direct search, ChatGPT recommendations, and reputation.
Practically, that means fewer thin articles chasing informational keywords, and more genuinely differentiated assets: original research, expert commentary, tools, and content that answers the messy, multi-step questions AI Mode encourages. Produce less, but make each piece something the model wants to quote and the human wants to open.
Sources
- AI Overviews, LLM, « collapse du funnel » : comment l’IA rebat les cartes du marketing — Blog du Modérateur
- AI Overviews : comment mesurer la visibilité de votre site web — Blog du Modérateur
- Google déploie AI Overviews et AI Mode en France — Blog du Modérateur
- Comment exclure votre site des AI Overviews de Google — Blog du Modérateur
- AI Overviews et AI Mode : le guide pratique pour optimiser votre contenu — Blog du Modérateur
- AI Overviews : ce que la presse française peut apprendre du déploiement au Royaume-Uni — Blog du Modérateur
