GEO maximizes the probability that a brand will be cited by AI (LLM). This visibility infrastructure replaces the click with the automated direct response.
For twenty years, the web functioned as a crossroads. An engine sent its users to sites, which monetized the audience through advertising, subscription, and commerce. This crossroads still exists, but it has moved. It is now inside the engine. The answers are formulated in the interface. Sources are cited unevenly. The clicks melt.
It’s not turbulence. It’s a diet change.
The word we put on it is new. We are talking about GEO (Generative Engine Optimization). I’m a little suspicious of the acronyms that appear at the same time as the slides. This one, however, designates something very concrete. It’s no longer about ranking well in a list of links. It is about being understood by a model, retained by it, cited in the response it produces. It is not an extension of SEO. It’s another job.
I will say it straight away: what I see taking shape today is not an optimization project. It is the recomposition of the economic contract of the web. The players who approach it from a technical angle will have already lost half of the stakes.
What fundamentally changes between SEO and GEO?
SEO was based on a simple promise: produce content, structure well, mesh well, let the engine sort. GEO responds to a different logic. The model does not classify, it reformulates. He doesn’t choose ten links, he writes a response. And to write it, he relies on sources that he considers reliable, corroborated, dated, signed.
This move seems technical. He is not. It changes what it means to exist on the web.
Before, we were fighting for rank. Today, we are fighting to be in the sentence. Before, the brand sent signals to an algorithm which rendered a binary verdict, you are there or you are not. Today, she sends signals to a model who writes about her, or not, based on what others have said about her. A major part of your generative visibility now takes place outside your site. It essentially plays out in the media, in the bases of authority, in the speeches of your leaders, in the mentions that your competitors do not make of you.
One figure is enough to measure this shift: around 84% of visibility in generative engines is today attributable to earned media, that is to say to what is said about you other than on your own site. Classic SEO rewarded mastery of the domain. GEO rewards the ecosystem that we build around us.
I see this shift happening from two very different angles depending on whether you are a publisher or a brand. Both play out in the same place, in the generated response. They have neither the same weapons nor the same horizons.
Why is GEO a business model topic for the media, not an SEO topic?
I come from this world. I have spent enough years working alongside editorial staff and management agencies to know what a drop in traffic of several tens of percent on the main referent represents. It’s not an adjustment. It’s the end of a cycle. The cycle where free distribution via an engine was enough to sustain a significant part of the ecosystem.
It must be said bluntly. A publisher who approaches GEO as a variation of SEO is missing the point. What is at stake now is not the place of its articles in a SERP. This is the contract that he negotiates, or does not negotiate, with those who reformulate his work.
Three areas seem to me to coexist in any serious writing today.
- Direct agreements with major AI players. The precursors, French and foreign, opened the way. These agreements do not solve everything, they are sometimes criticized, and are often opaque. But they set a principle: content has value, and this value must be paid for. The point of vigilance is that negotiation cannot be limited to the few leaders. The models are nourished by the entire web, and the challenge is to generally raise a principle of remuneration which also applies to medium-sized publishers.
- Usage-based revenue sharing programs. Some younger actors are experimenting with models where the editor touches based on quotes, direct visits and tasks entrusted to an assistant. It’s still imperfect, the measure is young, the volumes modest. But the idea is healthy: it pays for real use rather than speculation on a package.
- The industrialization of the relationship with crawlers. It is the most structuring project and the one about which we talk the least in the management committee. A publisher can today technically price AI access to its content, express machine-readable licensing conditions, block by default what must be blocked, open against compensation what must be. Bricks exist. They do not ask for absurd budgets. They are asking for a decision.
The real issue, for a decision-maker, is no longer to choose between one of these projects. It is to govern them together. A contractual license policy. A program participation policy for use. A technical access and pricing policy. Three policies, one subject, and one rule: no one else will do it for you.
Why is GEO, for a brand, a subject of story before being a subject of traffic?
The subject, on the brand side, arises differently. A brand, as a general rule, does not monetize the traffic of its showcase site. It monetizes preference, consideration, trust. And these three things are built today, largely, in interfaces that it does not control.
I see the same tipping point coming as in the media, but with a few years lag. Many marketing departments continue to buy media to occupy a SERP whose clicks are eroding. It’s useful, it’s not enough. The real question is no longer just about appearing, it is about being chosen as a reference by the model who writes the answer. An order of magnitude allows us to set the stakes: a brand cited in a generative response captures on average a third more organic clicks than a brand not cited on the same query. The quote becomes the new zero position. Those who aren’t there don’t lose a bit of traffic, they drop out of the conversation.
This needs to be worked on. It’s not magic. It’s slow, methodical, almost monotonous work. Four gestures make it up:
- Describe its entities precisely. The organization, its leaders, its products, its ranges, its commitments. A model doesn’t guess, he reads.
- Register them in the public databases that the models query. Wikidata for organizations and personalities, ISNI for authors, business repositories when they exist.
- Take care of your product data as you used to take care of your store records. Price, availability, specifications, returns policy, CSR commitments. What is missing is invisible.
- Feed a network of credible mentions. In the media, in speeches by identifiable experts, in sectoral studies. What others say about you now weighs more than what you say about yourself.
None of this is spectacular. All this decides, in the coming months, whether your brand is quoted or paraphrased without being named.
There is a wave coming behind that one, and it is more radical. That of the agents. An agent doesn’t click. He does not let himself be seduced by a banner. It reads structured data, compares, executes a transaction. To be chosen by an agent, you must be readable by an agent. What is not in the structured data does not exist for him. Payments made by autonomous agents have increased from barely 800 million dollars over the whole of 2025 to more than 2 billion in the first quarter of 2026 alone. The brands that have understood this are already working. Others will discover the subject when their AI-assisted sales start switching to a better-described competitor.
What are the three questions to ask the management committee?
They are simple, and I think they should be asked quickly.
- What is our exposure? How much of our traffic, visibility and acquisition today depends on requests that generative engines can absorb without returning a user to us? Is this share stable, or already falling? Many management teams discover, by seriously looking at their dashboards, that they have already started to lose ground without having expressed it.
- Where do we want to be on the GEO value chain? For a publisher, this means choosing between suffering, licensing, monetizing. For a brand, this means choosing between waiting to be cited by accident and methodically building the signals that make you so. This is not a technical question. It’s a question of position in your market.
- Who carries the subject internally? GEO intersects with editorial, structured data, technical architecture, brand strategy, legal affairs, finance, and sometimes commercial management when it comes to negotiation. Confining it to an SEO team is the best way to avoid it. A cross-functional referent, clear mandate, arbitrations reported to the COMEX. Otherwise the subject dies between two perimeters.
What is really going on behind the word GEO?
We can summarize all this as an optimization project. This would, again, be a short read. What is at stake is the reconstitution of a balance which has held for twenty years. Publishers produced, engines distributed, brands monetized, users clicked. This equation no longer holds. The engines respond. Publishers lose their free audience. Brands become invisible or pay to stay mentioned. Users click less and sometimes purchase through agents.
Everyone has a role to play, and a place to defend.
- Publishers who negotiate early, structure their access to AI, adopt new standards and sign non-exclusive agreements secure their margins for the next five years.
- Brands that invest today in their entities, their product data, their authority signals and their citation measurement gain a place that their competitors will take years to take back from them.
- Others will discover, by typing their own name in a generative response, that the available narrative about them was written without them.
I think we still underestimate this point. Generative visibility is not a budget that can be rekindled. It’s a position that we build. A field that sends no signal today will not easily catch up, in eighteen months, with a competitor who has started to structure its present.
AI hasn’t just changed the way users search. It changes the way in which markets are told, distributed and transacted. The companies that take the subject seriously in the coming months will not necessarily be the largest. They will be those who understand that being found, being cited and being executed by an agent are no longer the same thing. And that we must, from now on, work on all three.
