The rush towards generative engine optimization is based on a misunderstanding. This is not a new discipline to plug into a brand.
For several months, I have been receiving the same question, worded almost identically. A Shopify store founder wrote to me: “How do we get out into ChatGPT?” Implied: what is the manipulation, the adjustment, the piece of code to add to appear in the AI responses. The question is legitimate. The premise is false.
GEO is experiencing exactly the cycle that SEO experienced fifteen years ago: a wave of promises, an avalanche of “2026 guides”, tools that appear every week, and a stubborn belief that there is a technique to fool the machine. We’ve seen this movie before. We know the end.
What the data really says
When we look at serious analyzes rather than threads, the observation is regular, and a little destabilizing. Ahrefs cross-referenced 75,000 brands with millions of responses generated by ChatGPT, Google AI Overviews and AI Mode. The main conclusion is not a trick. This is a clarification.
Stacking pages doesn’t work: content volume doesn’t improve visibility in AI. Neither does mechanical optimization. What stands out are the brands that already exist as entities, those that are cited in comparisons, mentioned in articles, recommended on forums, present in the conversation well beyond their own site.
In other words: AIs do not create brand reputations. They reflect it.
This is a point that deserves attention, because it changes the entire logic. On Google, we could, for a time, compensate for a lack of notoriety with technology. Structure well, mesh well, target well, and reassemble. With generative engines, this shortcut closes. The model does not classify a page: it synthesizes what it understands about a brand from everything that is said about it. If few things are said, he is not inventing.
The problem is not GEO. This is what he makes visible.
This is where things get uncomfortable for some of the DTC brands I come across.
Many e-commerce stores have grown on a single fuel: Meta advertising. As long as the faucet flows, the number goes in. The brand looks like a brand. But much of that presence is rented, not owned. It disappears as soon as the budget is cut.
These brands are now discovering that they are not only fragile when it comes to paid acquisition. They are also absent from AI responses. And for a simple reason: we cannot be cited as a reference in a market where we have never built a reference. GEO does not create this hole. He brings it to light.
SEO was already doing this work, in fact. A brand that didn’t rank on anything, despite dozens of product pages, was already receiving the signal. We often preferred to attribute it to a technical problem, to a competitor “who was cheating”, to Google. AIs are less easy to blame. When a lesser-known competitor stands out in a ChatGPT response and you don’t, the technical explanation holds up poorly. What comes out is what the market says about each person.
Which does not mean that technology is useless.
I want to be specific, because the subject lends itself to shortcuts in both directions.
There are concrete things to do. Making its content extractable, structuring its pages, ensuring the consistency of its brand entity across the entire web, existing on the sources that the models really consult, it turns out that a significant part of ChatGPT’s citations overlap with the first results from Bing, which almost no one optimizes. These are real levers.
But these are downstream levers. They amplify existing notoriety. They don’t replace it. Working on your GEO without having built a brand is like polishing a reflection in the hope of changing the object that is reflected.
Order matters. The brand first. Then the structure. Finally, technology.
The real dividing line over the next two years
The risk, for brands that ignore this shift, is not a sudden drop in traffic. Google still sends the majority of visits. The risk is slower, and more difficult to recover.
It’s an erosion. As buyers ask their questions directly to an AI, “what is the best brand for They do not lose a ranking. They lose their place on the list before it is drawn up.
We experienced this dynamic between 2010 and 2015 on Google. Brands that structured early captured authority. Those who waited never really closed the gap. The window is opening again, moved by a few years and a notch more demanding, because we can no longer compensate for the deficit in notoriety with volume.
The good news is that it puts the right problem back in focus. The question is not “how do I hack ChatGPT”. The question is “why would an AI, or a human, recommend my brand over another”.
It’s a business question. She always was. Generative engines simply took away the ability to avoid it.
The brands that gain AI visibility will not be those that have found the best trick. They will be those who have accepted that GEO cannot be worked on the surface, because it only makes visible the work that has not been done below.
