More and more people no longer “search”: they ask. ChatGPT, Gemini, Perplexity, Claude. And instead of a list of ten links, they get a written answer built on two or three sources. For a business, the question changes in nature: it is no longer just about ranking well on Google, but about being the source a model draws on.

This shift has a name: GEO, for Generative Engine Optimization. It's not one more magic trick. It's the logical consequence of how these AIs produce their answers.

How an AI chooses what it cites

Most recent answers don't come from the model's “memory”. The system turns your question into one or more searches, retrieves pages, selects passages from them, then writes an answer, often with citations. In other words: to be cited, you first have to be found, then understood, then easy to reuse, at the exact moment the answer is being built.

Four conditions, not a formula

No provider publishes a recipe that guarantees you'll be recommended. You don't “win” a ranking: you increase a probability, measured across many questions. And that probability rests on four conditions that must all be met at the same time:

  • Be findable. AI crawlers must be able to explore your public pages, and nothing should block them by accident.
  • Be understood. The model must know unambiguously who you are, what you offer, to whom, and what evidence backs it up.
  • Be extractable. The information must be in the page's text, structured, dated and factual, not hidden behind JavaScript or drowned in slogans.
  • Be corroborated. Other independent sources must confirm that you exist and what you can do. Your site alone only makes claims.

Why traditional SEO is no longer enough

Two very concrete differences. First, many modern sites only display their content after JavaScript runs: a human sees a beautiful page, but a crawler that doesn't execute that code sees an empty shell. Second, a good Google ranking doesn't guarantee that you're the source an AI draws on: it may cite a page without naming the brand, or name a brand based on an external source.

On Google, you want to be found. In an AI, you want to be picked up. They're not the same thing, and you don't prepare for them the same way.

What really matters is clarity

An AI favors sources that make its job easier: a clear definition in the first few lines, dated and sourced facts, readable tables, a point of view of your own rather than an interchangeable summary. Google now says it explicitly: generic content that any model could produce is of no further interest. What counts is real experience and perspective.

The invisible trap

The most frustrating part is that you can have excellent content and still be invisible for a purely technical reason. A misplaced tag, rendering that depends on the browser, a simple server rule that sends crawlers a generic page, and all the editorial work vanishes. These problems can't be seen with the naked eye: you detect them by looking at what a crawler actually receives, without running any JavaScript. That's often where the gap hides between “our site is great” and “AIs never mention us”.

Where to start

  • Check what a crawler sees. The main content, title, text, contact details, must be present without running JavaScript.
  • Say clearly who you are. One page that answers “what is it, for whom, what evidence” in black and white is worth more than ten pages that circle around it.
  • Allow the right crawlers. AI search crawlers are often distinct from training crawlers: you can welcome the former without putting up with the latter.
  • Publish real substance. Facts, methods, measured cases that others will want to reuse, not volume produced just to fill space.

It's a technical project as much as an editorial one, and it's never settled once and for all: AI answers evolve quickly, and what's true this month deserves to be rechecked next month. But the underlying logic is stable: be findable, understandable, extractable and corroborated. That's exactly what we build, for our clients as well as for ourselves.

The right question is no longer “am I first on Google?” but “if someone asks an AI about my line of work, does my name appear, and is it accurate?”