A brand's own website accounts for only 5–10% of what AI-powered search engines read. The rest is written by others. McKinsey estimates that those who fail to adapt will lose up to half their traffic.

For twenty years, being found meant appearing on a list of links. That assumption is changing. According to Adobe Analytics, which tracks more than one billion visits to retail websites, traffic from AI assistants increased fivefold year-over-year in the first quarter of 2026. It remains lower in volume than traditional search, but not in quality: it converts 42% better than other channels.

Spain is no exception. IAB Spain’s eCommerce 2025 Study includes AI among the sources of information used before making a purchase for the first time: 13% of online shoppers use it to gather information, 9% use it as a direct search channel, and 60% say they feel comfortable incorporating it into their decision-making process.

In the marketplace sector, the shift now has official figures. Amazon reported in its fourth-quarter 2025 earnings that Rufus, its conversational assistant, was used by more than 300 million customers and contributed nearly 12,000 million dollars in annualized incremental sales. Behind the scenes, COSMO—a semantic graph that doesn’t search for keywords but instead infers the buyer’s intent—is at work. The practical result is a narrowing of the search results: where there used to be a results page with dozens of listings, there is now a single response with just a few.

That’s the blind spot, and it’s more troubling than it seems. According to McKinsey, a brand’s own website accounts for only 5% to 10% of the sources consulted by AI-powered search platforms; the remaining 90% consists of media outlets, reviews, forums, and third-party content. Put another way: almost everything a model knows about a brand has been written by someone other than the brand itself. Optimizing the brand’s profile is no longer enough when the conversation that feeds the model takes place outside the brand’s own channels.

The cost comes from traffic: McKinsey estimates that those who fail to adapt could lose between 20% and 50% of the traffic they currently receive from traditional search. And almost no one is monitoring this: only 16% systematically track their performance in AI-powered search. There is a lever: the research that coined the term GEO (Princeton, Georgia Tech, and IIT Delhi, KDD 2024) demonstrated that structuring content with data, cited sources, and direct answers increases the probability that a model will mention a brand by up to 40%; when applied to a catalog, this is what is now known as GEO optimization for AI search.

Carlos Martínez, CEO of Epinium, will address this shift at Tech Show Madrid / e-SHOW with the presentation “Your Brand Is Already Invisible to AI (and You Don’t Even Know It Yet)” (Wednesday, November 4, 12:00 p.m., Future of Commerce Auditorium). At FBA Show, he’ll showcase the other side of this same shift with a live demo: how to automate a marketing department by combining agents, skills, MCPs, and workflows. The underlying principle is the same: when the point of contact is no longer a person, both what needs to be published and who carries out the work change.