Is GEO different from AEO?

Not in practice. AEO came from featured snippets and voice answers; GEO was named in a 2024 paper on answers written by language models. Both mean writing pages that answer systems take their answers from. This site uses GEO and treats AEO as the older name.

Which of these terms matter most for a small business?

Four: the mention rate and share of voice, which say whether you are named and how much of the space you hold; citability, which says whether your pages can be quoted; and the crawler, because nothing else counts until one can read your site.

Where are the rest of the terms?

In the full glossary, which covers many more, each with a definition, where the term came from, and what it means for a business's pages. The cheat sheet links every term it lists to its entry there.

GEO cheat sheet: thirty terms, one line each

The practice

The practice: the terms, one line each, each linked to its full glossary entry.
Term	In one line
GEO	The work of making a business more likely to be named, cited or recommended in the answers AI assistants write.
AEO	An older name for the same practice, from the days of featured snippets and voice answers.
AI visibility	How often, and how prominently, the assistants name a business when asked what its customers ask.
Conversational search	Asking in ordinary language and getting a written answer, with follow-up questions read in context.
Zero-click search	A search that ends without the searcher visiting any page, because the answer was on the screen.

Measuring it

Measuring it: the terms, one line each, each linked to its full glossary entry.
Term	In one line
Visibility score	One number from 0 to 100 summarizing how often a fixed panel of questions names the business.
Mention rate	The share of answers, across the panel, that name the business at all.
Share of voice	The business's part of every business name given across the panel's answers.
Share of model	The share of one assistant's answers, across the panel, that name the business.
Prompt	The text given to a model to work from: a question, an instruction, a document.

Being quoted

Being quoted: the terms, one line each, each linked to its full glossary entry.
Term	In one line
Citability	The quality of a passage that lets an assistant lift it as an answer and still credit whoever wrote it.
Answer page	A page built around one question, with the full answer near the top.
Source citation	The link an assistant attaches to a statement, pointing at the page it came from.
Brand mention	The business's name in text it did not write, linked or not: a review, an article, a thread, an answer.
Featured snippet	The passage Google shows above its results, chosen by Google, which a site can only decline.

The markup

The markup: the terms, one line each, each linked to its full glossary entry.
Term	In one line
Structured data	Machine-readable markup stating what a page is about: which business, where, open when, selling what.
JSON-LD	The way of writing that markup as JSON, which Google recommends.
LocalBusiness schema	The schema.org type for a business with a place, and the markup describing its address, phone and hours.
Review schema	Markup attaching a rating to what it rates; a business's own reviews of itself earn no stars on its own site.
FAQ schema	FAQPage markup listing a page's questions and answers; Google stopped showing its rich result in May 2026.
Entity	One identifiable thing, such as a business, a person or a place, that a machine can tell apart from others.

How the assistants work

How the assistants work: the terms, one line each, each linked to its full glossary entry.
Term	In one line
Large language model	A model trained on large amounts of text to predict the next token, which lets it write answers.
Retrieval-augmented generation	Answering from documents fetched for the question, rather than from memory alone.
Knowledge cutoff	The date a model's training data stops; anything later is news to it.
Hallucination	A statement presented as fact with nothing in the training data or the retrieved pages to support it.
Query fan-out	Turning one question into several searches of the system's own.
Agentic search	An assistant searching in several steps, reading and searching again before it answers.

The web underneath

The web underneath: the terms, one line each, each linked to its full glossary entry.
Term	In one line
Crawler	The program that fetches pages for an index or an answer; each assistant names its own.
llms.txt	A proposed file listing a site's key pages for language models; a proposal, not an adopted standard.
E-E-A-T	Experience, expertise, authoritativeness and trust, the qualities Google's raters are asked to judge.

The cheat sheet: thirty terms, one line each