ChatGPT and Gemini now decide which brands consumers see

When a Mexican consumer asks an AI assistant which car to buy, which bank to choose for a first credit card, or where to buy clothes, they no longer get a list of hundreds of results the way a traditional search engine would show. They get three or four names — and whether or not a brand appears in that answer has become a new battleground.

That is the conclusion of the first geoSurge México Visibility Index, a study built from 14,000 responses by ChatGPT and Gemini to 700 questions asked in Spanish. The analysis covered seven industries — automotive, banking, electronics, video games, retail, telecommunications and travel — and compared how recommendations change when the AI has internet access versus when it answers only from what it memorized during training.

Toyota, Suburbia and Telcel dominate the answers

In the automotive sector, Toyota remains the most recommended brand whether the AI model answers purely from memory or searches the internet, according to the study. Telecommunications shows a similar pattern: Telcel keeps a dominant position in both scenarios.

The most revealing case is retail. When ChatGPT answers using only the data it was trained on, Amazon appears in 72% of responses. But as soon as the model has internet access and can pull up-to-date information, that presence drops to 16%, and Suburbia takes the top spot instead. In banking and electronics, fintech company Klar ranks among the brands most recommended by both assistants.

The study also measured how much weight the model's memory carries against live search: between 79% and 87% of brand-mentioning queries included companies that were already among the top ten options stored in the model's memory, a share that exceeds 97% in telecommunications. And when the model itself decided to search for a specific brand online, that brand ended up appearing in the final answer 89.6% of the time on ChatGPT and 96.9% of the time on Gemini.

The models don't treat questions about Mexican banks as an open investigation: they arrive with a short list already formed. The one who is named competes; the one who isn't doesn't exist.

From optimizing for search engines to optimizing for AI

What the Mexican study documents is part of a wider shift that already worries major consulting firms and enterprise software makers alike. In the United States, IBM recently presented a twelve-part playbook at the Adobe Summit conference to help brands survive what the industry calls generative engine optimization, or GEO. According to Alexis Zamkow, IBM's global lead for marketing transformation, as much as 75% of the visibility traditional search engines generate today could shift to AI agents within the next two years.

"These machines are disintermediating the brand experience," Zamkow warned during the presentation, as reported by trade outlet Search Engine Land. The firm argues it is no longer enough to simply be mentioned by an AI assistant: the goal is to be cited as a trusted source, which requires clear content, structured data, and consistent presence across the web — not just on the brand's own site. According to that same reporting, 85% of the mentions cited by generative AI engines come from external domains — forums, reviews, social media and news outlets — rather than a company's official website.

• Keep a website updated and easy to read, for people and AI crawlers alike.
• Publish clear content that explains in a few words what the brand offers and to whom.
• Check that no technical setting is accidentally blocking AI assistants' crawlers.
• Build a presence in reviews, forums and outside media outlets, which now generate more citations than a brand's own site.

For small and mid-sized businesses, the warning cuts both ways. Francisco Vigo notes that larger companies hold a structural advantage because they generate far more information online, which biases AI models toward already-established brands. According to Vigo, small businesses now face triple pressure: competing on traditional search engines, on social media, and on visibility to artificial intelligence, all at once. On IBM's side, Zamkow was just as blunt: when an executive asks why their brand doesn't show up in an AI recommendation, the answer is no longer a problem for the SEO team — it is a decision made at the leadership level.

For a business anywhere, this means competing no longer depends only on paying for search ads or tending to social media: it depends on whether ChatGPT and Gemini can find, understand and trust its information when a customer asks where to buy. And the window to react is shrinking: as AI moves toward what the industry calls "agentic commerce" — assistants capable of completing a purchase without human input — failing to show up in an AI's answer will stop being a visibility problem and become, directly, a lost sale.