Inteligencia Artificial
Two years ago, someone who needed to protect their company's data opened Google, typed «cybersecurity company near me» and worked through the first few results. Today more and more people open ChatGPT and type something else entirely: «I run a twenty-person company, I handle customer data and I don't know where to start protecting it». And they read the answer.
The difference isn't the tool. It's that in the second case nobody sees a list of websites. They see a paragraph with a recommendation. And if your brand isn't inside that paragraph, as far as that person is concerned you simply don't exist.
GEO stands for Generative Engine Optimization: optimising so that generative engines —ChatGPT, Perplexity, Google's AI summaries, Copilot— mention and cite you when they answer someone.
SEO works to get you into a list of links. GEO works to get you inside an answer. It sounds similar and it isn't. In ten results there are ten opportunities and the user picks. In a generated answer there are two or three mentions, the model picks them, and whatever falls outside doesn't drop a position: it doesn't appear at all.
For twenty years the deal with the search engine was implicit: you published useful content and it sent you visitors. That deal is breaking down. When the AI answers directly and answers well, the visit never happens, because the user already got what they needed without leaving the chat.
What does arrive is different, and generally better: people who already received your name as a recommendation from an assistant they trust. They arrive with the decision half made. Less traffic, more intent.
And there's an uncomfortable detail worth facing head-on. In ten blue links there was room for everyone: the seventh result still got something. In a three-sentence answer, there isn't. Whoever isn't mentioned isn't anywhere.
Models don't cite pages that «rank well». They cite pages that answer the question clearly, completely and verifiably. That pushes content in a very specific direction: less text written to please an algorithm and more text written to resolve a person's doubt, with concrete data, figures where they exist and statements that can be lifted and quoted as they are.
It also changes where you need to be. A classic search engine reads your website; a generative model also feeds on forums, directories, press, reviews and comparisons. Your reputation outside your own domain has gone from being a bonus to being raw material.
And it changes the technical side, though less than you'd think: structured data done properly, clean HTML, content that isn't hidden behind JavaScript. Nothing exotic. The usual, but now with a different reader on the other side.
There's one last thing almost nobody does yet: measuring. Asking the assistants exactly what a customer of yours would ask them, and noting whether you show up, what words they describe you with and which competitors they place you next to. There is no Search Console for this. The tracking is manual and repetitive, and that's precisely why almost nobody has it.
GEO doesn't replace SEO, it leans on it. Models feed largely on what is already indexed and on the same authority signals as always. If you don't exist for Google, it's very hard to exist for ChatGPT. Anyone framing GEO as a replacement for SEO hasn't understood where the information comes from.
There are no tricks either. There's no magic keyword and no way to buy a mention. Nobody sells you a slot inside a ChatGPT answer; if someone offers you one, you already know what they're selling.
And measurement is still immature. Answers change between users, between countries and between weeks, because the models get updated. Anyone promising you «position 1 on ChatGPT» is describing something that doesn't exist: there are no positions there, there are mentions. The honest approach is to measure trend, not ranking.
Finally, it doesn't weigh the same for everyone. If you sell services, if your customer compares before deciding, or if the decision is complex and expensive, GEO is already affecting you even if you aren't watching it. If you live off impulse purchases, it matters considerably less.
With the boring part, as usual. Writing content that answers real questions instead of chasing keywords. Getting the technical house in order. Looking after what is said about you off your own site. And building a set of key questions to review once a month to see whether you appear and how.
None of that produces results next week. It compounds, just like SEO: the work from month one is still paying off in month twelve. The difference is that today almost nobody is doing it, and that window won't stay open for long.
For twenty years the question was «how do I come first on Google?». Today's question is a different one, and considerably more uncomfortable: when an AI recommends someone a company like yours, why would it name you?