the plain-English glossary of AI search

The AI search conversation is drowning in acronyms, and half the confusion in this field is vocabulary. Here is every term worth knowing, defined plainly, one at a time, so the rest of the conversation makes sense. Each definition stands alone, which is the point.

SEO, search engine optimization, is the craft of being found and ranked by search engines, and it remains the foundation everything else stands on.

GEO, generative engine optimization, is the craft of being cited and quoted inside AI-generated answers. The term comes from a 2024 academic paper that first measured which content changes earn more AI visibility.

AEO, answer engine optimization, is an older cousin of GEO focused on winning direct answers, and in practice the industry now uses GEO and AEO almost interchangeably. The work is identical: be the clearest, best-corroborated answer.

AI Overviews are Google's generated answers at the top of many results pages, drawn from indexed pages and linked to their sources.

An answer engine is any system that responds with a written answer rather than a list of links: ChatGPT with search, Perplexity, Google's AI experiences, and whatever ships next.

A citation, in this context, is an answer engine crediting a source page for part of its answer, usually with a link. Citations are to AI search what rankings are to classic search: the visible score.

Retrieval, sometimes written RAG for retrieval-augmented generation, is the pipeline where an assistant searches, reads what it finds, and writes an answer grounded in those documents.

Query fan-out is what happens to your question inside that pipeline: it becomes several searches covering the angles of the question, which is why one page rarely wins alone.

A passage is the few-sentence chunk of a page that retrieval systems actually score and quote. Pages are containers; passages compete.

An entity is a thing the machine tracks as distinct, your business among them, assembled from every consistent mention of you across the web.

A knowledge graph is the map of those entities and their relationships that search systems maintain, and being well-defined in it is what lets an engine recommend you with confidence.

Structured data, or schema markup, is machine-readable labelling on your pages declaring what things are: an article, an organization, a person, a question and its answer.

E-E-A-T, experience, expertise, authoritativeness and trust, is Google's published framework for judging source quality, and a fair summary of what every answer engine approximates.

llms.txt is a small file some sites publish to describe themselves to AI systems in plain text. Cheap to add, honest to describe as promising rather than proven.

Corroboration is the degree to which the wider web agrees with what you say about yourself, and it is quietly the strongest signal of the lot.

Vocabulary shifts faster than practice in this field, and new coinages arrive monthly, most of which rename something on this page. When you meet a new acronym, translate it back to the plain question underneath: does this help a machine find, understand, trust or quote us? If it does none of those, it is a word, not a strategy.