GEO (Generative Engine Optimization) is the set of practices aimed at making a business, a brand or a piece of content visible in AI-generated answers: ChatGPT, Claude, Gemini, Perplexity, and the AI answers of search engines. Where SEO works on a website's position in a results page, GEO works on a name's presence in a written answer.
Why it became a topic
A growing share of buying journeys starts with a question asked to an AI: "which tool for...", "which company would you recommend in...", "what to think of...". The answer is not a list of ten links: it is a text that names a few companies. Either you are part of it, or you do not exist in that journey; there is no "page 2" of an AI answer.
This shift does not remove SEO: generative engines lean heavily on web search. It adds a layer on top, with its own rules and its own measurement. Advertising arrived through a different door: since February 2026 ChatGPT shows a sponsored slot under the answer, sold at auction and kept apart from the cited sources.
The terms in circulation (and what they cover)
The vocabulary of the discipline is still settling; in practice, these expressions cover the same reality, with nuances of angle:
- GEO (Generative Engine Optimization): the umbrella term, modeled on SEO. The tools that measure it are compared, on what a year of measurement costs, in our guide to GEO tools.
- AEO (Answer Engine Optimization): the same topic seen through the lens of "answer engines"; emphasizes the question-answer format. The tools that measure it are compared, on their coverage and their prices, in our guide to AEO tools.
- LLM SEO, AI SEO: common shorthand variants, especially in content; same practices.
- AI visibility: the measurable fact (appearing or not in the answers), independent of techniques.
- AI share of voice: your relative weight in the answers, compared with the other names cited on the same questions.
GEO and SEO: what really changes
- The unit of measurement. SEO measures a position in an ordered list. GEO measures a presence in generated text: your name is cited or it is not, with more or less prominence. There is no "position 7" in an answer.
- Variability. A results page is relatively stable from one query to the next; a generated answer changes from one run to the next. GEO measurement therefore relies on repetitions and is expressed as a citation rate, never as a one-off observation.
- The plurality of engines. Our weekly measurements have shown it since June 2026: on the same question, each engine has its own podium of cited names. Visibility is measured engine by engine.
- The role of sources. Generative engines cite and synthesize third-party pages: comparison articles, directories, press, reference content. How much those sources weigh against your own site is read question by question: on a category-discovery question, third parties carry the bulk of the citations; on a request for a recommendation in a trade and a city, the professional's own site carries 61.9% and directories 30.4%. And those sources are not only written: on two of the four engines, the single most cited source is YouTube.
What GEO measures
- The citation rate: the share of answers where your name appears, over a panel of questions asked continuously, with repetitions.
- Share of voice: your weight against the other names cited on the same questions.
- Cited sources: the pages engines draw their answers from; that is the work list of any GEO effort. How many an answer carries, and whether it carries any at all, depends on the engine.
- Reputation: the tone and the restated facts when the AI is asked directly about your name, to be distinguished from spontaneous recommendation.
All of it is observed over time: a one-off measurement ages quickly, the longitudinal series is where the value lies. The full method is described on our Method page.
Where to start
With measurement, not promises: establish your baseline (where you appear, at what rate, fed by which sources), work on what can be observed, re-measure. A brand the engines do not know yet starts the other way around: write first, measure from zero. Our Tracking page describes the measurement protocol; our Method page sets out the levers, step by step. If the choice of instrument comes first, AI visibility tools are compared on published facts. What can be steered is presence in the sources, and measurement of the effect. Which sources? Our survey of the site profiles behind AI answers gives the picture engine by engine, and our study on backlinks weighs what the link still buys there.