GEO glossary

GEO, AEO, LLMO, llms.txt, GPTBot, grounding, share of voice. The terms of answer-engine optimisation, explained without jargon.

GEO: GEO — Generative Engine Optimization — is the work of making a brand citable by AI answer engines such as ChatGPT, Perplexity and Gemini.; It differs from SEO in its object: SEO competes for ten links on a results page, GEO competes for three names inside a written answer. There is no average position and no CTR — either the brand is named, or it does not exist in that answer.
AEO: AEO — Answer Engine Optimization — is the same work as GEO, under a name that insists on what changed: the user gets an answer, not a list.
LLMO: LLMO — Large Language Model Optimization — is the variant that stresses the model rather than the engine, used mostly when discussing training rather than search.
llms.txt: `llms.txt` is a file at a site's root that summarises, in plain text, what the site is and where the content that matters to a language model lives.; It is not a mandatory standard and no engine requires it. It is cheap to produce and, when present and well structured, saves the model from guessing based on navigation.
GPTBot: GPTBot is OpenAI's crawler that collects content for model training. Blocking it in `robots.txt` prevents the site from feeding future versions of ChatGPT.; Not to be confused with `OAI-SearchBot` and `ChatGPT-User`, which serve real-time search. Blocking the latter is what removes a brand from today's answers; blocking GPTBot affects what the model knows tomorrow.
Grounding: Grounding is the web search a model performs before answering, to base its answer on current sources rather than on training memory.; It is what makes measurement possible and what makes it expensive: providers charge for each search separately, and one request can trigger several.
Share of voice: Share of voice, in AI visibility, is the share of answers across a set of questions in which a brand is named, compared against competitors.
robots.txt: `robots.txt` is the file telling each crawler what it may or may not visit. It is the first place to look when a brand never shows up in answers.; A `Disallow: /` aimed at `User-agent: *` also blocks AI crawlers — and is usually inherited from a staging environment nobody revisited.
Dados estruturados: Structured data is JSON-LD markup telling a machine what a page is — a company, a service, a question with an answer — instead of forcing it to infer from the text.; `Organization` and `LocalBusiness` identify the entity; `FAQPage` is the format models extract most, because it already comes in the question-and-answer shape they produce.
Citação e menção: A mention is the brand name appearing in the answer text; a citation is the engine pointing at the site as a source. There can be mentions without citations, and citations without traffic.