AI Visibility Glossary: GEO, RAG, Schema and Entity Terms

The practice of making information clear, accessible and useful to systems that provide direct answers rather than only lists of search results.

A system that interprets a question and constructs a direct response from model knowledge, retrieved sources or a combination of both.

Automated software used to discover and retrieve web content for AI search, model training, live answers or other machine-processing purposes.

The extent to which AI-powered search and answer systems can discover, understand, verify and accurately represent a business, person, service or subject.

A visible reference or link identifying a source used to support an AI-generated statement. Citation does not necessarily mean endorsement.

Independent or supporting evidence that confirms an important claim, identity, qualification, relationship, service or business fact.

The ability of an authorised search or AI crawler to reach and retrieve website content without being blocked by technical faults, firewalls, directives or authentication barriers.

The qualities Google asks its search quality raters to consider when assessing whether content and its source appear reliable and useful. It is not a single ranking score.

A numerical representation of text, images or other information that allows an AI system to compare meaning, context and similarity.

A distinct, identifiable thing—such as a business, person, place, service, product or organisation—that can be described and connected to other things.

The deliberate organisation of entities and relationships so that machines can connect a business with its people, services, content, locations and supporting evidence.

The process of determining whether different names, records, pages or references describe the same real-world person, business or object.

Work intended to improve how accurately and confidently generative search and answer systems can discover, interpret and use published information.

Search that uses generative AI to construct a summarised or conversational response, often using information retrieved from multiple sources.

Connecting an AI-generated response to retrieved data, documents or verifiable sources rather than relying only on the model’s internal statistical knowledge.

An OpenAI web crawler used to access publicly available website content for purposes described in OpenAI’s published crawler documentation.

An output that appears plausible but contains invented, unsupported or materially incorrect information.

A JavaScript-compatible format commonly used to publish structured data without changing the visible wording or layout of a webpage.

A structured network of identifiable entities and the relationships between them, allowing information to be understood as connected meaning rather than isolated text.

A machine-learning model trained on large collections of language data to interpret prompts and generate text or other structured outputs.

A proposed website file presenting selected information and important resources in a form intended to be convenient for language models. It is not a universal crawling or indexing standard.

The traceable origin, authorship, ownership and history of a piece of information, claim, document or digital asset.

A method in which an AI system retrieves relevant information from external sources and uses it as context when generating a response.

The process of locating and selecting information considered relevant to a query before an answer is generated or displayed.

A file placed at the root of a website that gives compliant web crawlers instructions about which areas they may or may not crawl.

Structured vocabulary added to a webpage to clarify the entities, properties and relationships described by its visible content.

Search that attempts to understand meaning, context, relationships and user intent rather than matching only the literal words in a query.

The degree of confidence a system may place in a source based on its relevance, expertise, reputation, consistency and supporting evidence.

Information organised using a predictable format and vocabulary so software can identify its meaning and relationships more reliably.

A database designed to store and search numerical embeddings, commonly used to retrieve information according to semantic similarity.

A method of finding information by comparing numerical representations of meaning rather than relying solely on exact keyword matches.