GEO and AEO glossary

Terms A to Z

AEO (answer engine optimization)

The practice of making your content the direct answer to a question. The term predates today's AI assistants (it was used for featured snippets and voice assistants) and is now often used interchangeably with GEO.

AI Mode

Google's conversational search experience, which answers questions with a generated response and links to sources. Google's documentation describes how its AI features relate to Search; check the current wording, as it is updated.

AI Overview

A generated summary that Google can show at the top of some search results, with links to supporting pages. Whether one appears depends on the query.

Answer engine

A product that responds to a question with a written answer instead of only a list of links. ChatGPT search, Perplexity and Google's AI features are examples.

Canonical URL

The preferred address for a page, declared with a rel="canonical" link. It tells crawlers which of several near-identical URLs to treat as the original.

Citation

A visible reference to a source that supported an AI-generated answer, usually a link. A source can influence an answer without being cited.

Crawler (user-agent)

An automated program that fetches web pages and identifies itself with a name called a user-agent, such as GPTBot or Googlebot. Site owners can allow or disallow specific crawlers in robots.txt.

Entity

A distinct, identifiable thing such as a company, person, product or place. Search and AI systems try to work out which entities a page is about and what is true of each.

Evidence tag (OBSERVED, INFERRED, NOT_CHECKED)

Venoh's labelling of each finding by how it was determined: seen directly in the fetched page, a reasonable conclusion from it, or outside what was examined. See the methodology page.

FAQPage

A schema.org type for a page that contains a list of questions with their answers. It should be used only when those questions and answers are visible on the page.

Generative engine

A search or answer product that uses a generative language model to write its responses. The term was popularized by the 2023 research paper that named GEO.

GEO (generative engine optimization)

The work of improving how accurately and how often a company appears in answers generated by AI systems, by making its information easy to find, understand and cite.

Grounding

Supplying a language model with retrieved source material so its answer is based on that material rather than only on what it learned in training.

Hallucination

A statement generated by a language model that sounds plausible but is false or unsupported by any source. Clear, specific, consistent facts on your own pages reduce the room for errors about your company.

JSON-LD

A format for writing structured data as JSON inside a script tag, commonly used to publish schema.org markup. It is the format most site owners use for structured data.

Knowledge graph

A structured database of entities and the relationships between them. Search engines maintain them to connect facts about people, companies and things.

Large language model (LLM)

A neural network trained on very large amounts of text to predict and generate language. It is the component that writes the wording of an AI answer.

LLMO (large language model optimization)

A broad label for influencing what language models say about a subject, sometimes including what they learn during training. Usage is inconsistent between writers.

llms.txt

A proposal for a Markdown file at the root of a site that gives language models a curated index of useful pages. It is not an established standard and support by AI providers is not guaranteed.

Mention

An appearance of your company's name in an AI-generated answer, with or without a link. Mentions and citations are tracked separately.

Organization schema

A schema.org type that describes a company: its name, website, description and contact details. It is one of the most useful basics for stating who you are.

Passage

A short section of a page, often a heading plus the text beneath it. Retrieval systems frequently rank and extract passages rather than whole pages.

Prompt

The question or instruction a user gives an AI system. In measurement, a fixed set of representative prompts is run repeatedly to track how a brand appears.

Query fan-out

A technique where a system splits one question into several related searches and combines the results. It means a page can be retrieved for a query the user never typed.

RAG (retrieval-augmented generation)

An approach in which a system first retrieves relevant documents and then has a language model write an answer using them. Most search-connected AI assistants work in a way that resembles RAG.

robots.txt

A plain text file at the root of a site that tells crawlers which paths they may or may not fetch. It is a request that well-behaved crawlers honor, not a security control.

Schema.org

A shared vocabulary of types and properties (Organization, Product, FAQPage and many more) used to describe the content of web pages in a machine-readable way.

Share of voice (AI)

A measure of how often a brand is mentioned or cited in AI answers to a set of prompts, compared with other brands. Because answers vary, it must be sampled repeatedly to mean anything.

Sitemap

An XML file listing the pages of a site you want crawled, with optional last-modified dates. It helps crawlers discover pages but does not guarantee indexing.

Structured data

Markup that states facts about a page in a form software can read directly, typically schema.org types written as JSON-LD.

Zero-click search

A search in which the user gets what they need from the results page or the generated answer and never visits a website.