AI search dictionary

AEO is editorial and technical work that helps a page answer a specific question clearly and with evidence. The term describes a practice, not a distinct algorithm, file, or schema.org type.

An AI citation is an observable source link or attribution in a particular AI answer. It is an answer-level sample outcome, not a promise that a page will always appear.

AI mention is the observed appearance of an entity in answer text or its structured presentation. It is not a ranking position or proof of a click. A mention, link, citation, and recommendation are separate fields. One answer can contain all four, but none substitutes for another. A nearby source is not necessarily the cause of the wording. It is especially important to separate the displayed mode from assumptions about the internal pipeline.

An AI recommendation is an answer in which a system suggests an entity as suitable for the user’s task. It should be separated from a neutral mention and advertising. A mention, link, citation, and recommendation are separate fields. One answer can contain all four, but none substitutes for another. A nearby source is not necessarily the cause of the wording. It is especially important to separate the displayed mode from assumptions about the internal pipeline.

AI SEO applies durable search principles to surfaces with generative answers: accessibility, indexability, useful content, clear entities, and verifiable sources. Google does not require separate mandatory “AI markup.”

AI visibility is the observed appearance of a brand, page, or entity in an AI-system answer. A mention, recommendation, citation, and referral are different outcomes; one bot visit or one favorable answer does not measure overall visibility.

AIO is a broad, unsettled label for optimization involving AI interfaces. Before using it, state whether it means search, answers, a product feature, advertising, or process automation.

Answer engine is an editorial term for a system that receives a question, retrieves context, and forms an answer. It is not a universal official product class or a guarantee of sources.

Answer model is a working label for the generative part of an answer engine that turns a query and available context into answer text. It does not mean a researcher knows the exact model: an interface may route between models and apply post-processing. Reports should state the known model, mode, and confidence, leaving unknowns unknown.

Content freshness means that the facts on a page are still accurate as of the stated date. Changed facts, sources, and limitations need a real review. Updating the displayed date without changing the content does not make a page fresh.

A crawler is a bot that automatically opens web pages. Its software identifier, called a User-Agent, can be copied, so identity needs another check. Even a verified visit proves only access to the page, not that the page was indexed or used in an answer.

An entity is a distinguishable thing about which stable facts and relationships can be represented, such as an organization, person, product, place, or concept. A site should name it consistently and state its relationships explicitly, but that does not guarantee an entry in another system's knowledge graph.

GEO is the practice of making public information accessible, understandable, and verifiable for generative search systems. It is neither an official platform standard nor a way to guarantee a mention or link.

Grounding means connecting an answer to evidence that lets a reader check specific claims. A list of links beside an answer is not enough: it should be clear which source supports which claim.

Indexability means a page is technically eligible to be added to a search engine's index: the page is accessible and does not forbid indexing. It does not guarantee that the page will actually be indexed, rank highly, or appear in an AI answer.

A knowledge graph stores facts as a network of things and the relationships between them, such as “a company owns a product.” Site markup can help describe those relationships, but it does not prove that a search or AI system accepted them.

Query fan-out means splitting one complex request into several narrower searches and then combining the results. A long answer or a large link list does not by itself prove that this happened.

RAG (retrieval-augmented generation) is a method in which a system first finds useful passages in an external collection, then uses them to draft an answer. It does not automatically mean open-web search, fresh data, accurate facts, or visible source links.

Retrieval is the stage at which a system finds and selects documents or passages that may help answer a request. It is not the same as a bot visit, inclusion in a search index, or showing a source link to the user.

robots.txt is a public file where a site owner tells named bots which sections they may or may not visit. It is not access protection and does not guarantee removal of a page from a search index.

A search mode is the selected way of processing a query: classic results, an AI answer, a web-enabled chat, or another explicitly named mode. A mention, link, citation, and recommendation are separate fields. One answer can contain all four, but none substitutes for another. A nearby source is not necessarily the cause of the wording. It is especially important to separate the displayed mode from assumptions about the internal pipeline.

A search surface is the specific interface and interaction mode in which a search result is observed: for example, a classic results page, an AI answer, a chat, or a map. One platform can expose several surfaces with different sources and rules. A metric should therefore record the surface, mode, locale, and date instead of saying only ‘visibility in Google.’

A sitemap is a technical list of a site's primary page addresses that helps a search engine discover them. It may also state when a page changed substantially. A sitemap does not guarantee a crawl, indexing, a high position, or inclusion in an AI answer.

Source attribution is a visible connection between a claim in an AI answer and a named source, such as a label, source card, or link. It states provenance but does not by itself prove that the source supports every word of the answer.

Structured data is a technical label in page code that explicitly identifies details such as the title, date, and author. JSON-LD is one common format. This markup can help a system understand the page, but it does not guarantee indexing, an enhanced search result, or a source link in an AI answer.