GEO Expert
Let’s talk

The GEO glossary

38 terms used in generative engine optimization and AI search, each defined in two or three plain sentences, with links to related terms and the tools that put them into practice.

Sai Siddesh

Terms from A to Z

#AI Mode

A Google Search mode that answers a question with a generated response and supporting links, using query fan-out to run several related searches behind the scenes. Google says pages need no special optimization to appear, beyond being indexed and eligible for a snippet.

Related: AI Overviews, Query fan-out

#AI Overviews

Google's generated summaries that appear above regular results for some searches, with links to the sources they draw on. A page is eligible as a supporting link if it is indexed and eligible to show a snippet in Google Search.

Related: AI Mode, Snippet controls

#Answer engine

Any system that responds to a question with a written answer rather than a list of links, such as ChatGPT, Perplexity, Gemini, Claude, Copilot or Google's AI features. Generative engine means the same thing in the research literature.

Related: Generative engine

#Answer engine optimization (AEO)

An industry name for the same work as generative engine optimization: making a business or page the source an AI system names and cites. The two terms are used interchangeably.

Related: Generative engine optimization (GEO)

#Answer-first content

Writing that states the direct answer to the reader's question in the opening sentences, within about sixty words, before giving reasons and detail. It is the format assistants find easiest to quote. Go deeper →

Related: Passage, Self-contained sentence

#Citation

A link or named reference to a source inside an AI-generated answer. Being cited is distinct from being named: an assistant can mention a business without citing its website, or cite a directory page about it instead.

Related: Citation share, Named share

#Citation share

The share of answers on a fixed question panel that cite one of the business's own pages as a source. It shows whether the business's own content, rather than third-party pages about it, is doing the work.

Related: Named share, Question panel

#Consideration set

The small group of businesses an AI system treats as plausible answers to a question. For local and professional services it is usually stable across differently worded versions of the same question.

Related: Named share

#Corroboration

Independent sources stating the same facts about a business: its name, location, people and services. Assistants look for agreement between sources before naming a business, so corroboration in places the business does not control raises confidence.

Related: Entity reconciliation, NAP consistency

#Crawl access

Whether the crawlers that feed AI search can reach and read a page. It depends on robots.txt rules, server and firewall settings, and whether the main text is present in the HTML without running scripts. Go deeper →

Related: Robots.txt, Server-side rendering

#Entity

A distinct, identifiable thing such as a business, person, place, service or concept. Search engines and AI systems reason about entities and the relationships between them, not just words on pages.

Related: Entity reconciliation, Knowledge graph

#Entity reconciliation

Making every public record of a business agree, so systems can confirm they all describe one organisation. It covers the exact name, address, phone number, people, services and licence details across the website, profiles, registers and directories.

Related: NAP consistency, Corroboration

#Fresh session

A new chat or private browser window with no history, used when testing what an assistant says. Earlier messages in a conversation change later answers, so measurement must start clean every time.

Related: Question panel

#Generative engine

A system that retrieves sources and writes a single synthesised answer, as defined in the 2023 GEO paper by Aggarwal and colleagues. ChatGPT search, Perplexity, Gemini and Google's AI Overviews are examples.

Related: Answer engine, Generative engine optimization (GEO)

#Generative engine optimization (GEO)

The practice of making a business, person or page more likely to be retrieved, trusted and cited by AI systems that write answers. The term was introduced in a 2023 paper published at KDD 2024. Also spelled optimisation. Go deeper →

Related: Answer engine optimization (AEO), GEO-bench

#GEO-bench

The benchmark of real-world queries across many domains built for the original GEO research, used to test how rewriting content changes its visibility in generated answers. Adding statistics and quotations performed best; keyword stuffing did not help.

Related: Generative engine optimization (GEO)

#GPTBot

OpenAI's crawler for content that may be used to train its foundation models. Blocking it in robots.txt opts a site out of training; it does not control whether the site appears in ChatGPT search, which is governed by OAI-SearchBot.

Related: OAI-SearchBot, Robots.txt

#Grounding

Tying an AI model's answer to retrieved sources rather than to what it memorised during training. Grounded answers can cite where their facts came from, which is what makes citation possible.

Related: Retrieval-augmented generation (RAG)

#Hallucination

A confident statement by an AI system that is false, such as a wrong address, a service the business does not offer, or a business that has closed. Contradictory or stale public records make hallucinations about a business more likely.

Related: Entity reconciliation

#IndexNow

An open protocol that lets a site notify participating search engines, including Microsoft Bing, the moment pages are added or changed, by posting the URLs together with a key file hosted on the site.

Related: Crawl access

#Knowledge graph

A database of entities and the relationships between them, used by search engines and AI systems to answer factual questions. Consistent public records, with structured data as a supporting signal, help a business be represented accurately in it.

Related: Entity, Structured data

#Large language model (LLM)

A model trained on large amounts of text to predict and generate language. The assistants people ask for recommendations are built on LLMs combined with search and retrieval.

Related: Grounding

#llms.txt

A proposed markdown file at a site's root that gives AI systems a short summary of the site and links to its most useful pages, proposed by Jeremy Howard in September 2024. It is a voluntary convention, not a guaranteed input to any AI product. Go deeper →

Related: Crawl access

#Named share

The share of answers on a fixed question panel that mention a business at all. It is the headline measure of AI visibility, compared against competitors measured on the same run.

Related: Citation share, Question panel

#NAP consistency

Using exactly the same business name, address and phone number everywhere a business is listed. It is the simplest part of entity reconciliation and a common reason businesses are left out of answers.

Related: Entity reconciliation

#OAI-SearchBot

OpenAI's crawler for ChatGPT's search features. OpenAI says sites that block it will not appear in ChatGPT search answers, and recommends allowing it in robots.txt.

Related: GPTBot, Robots.txt

#Passage

A short span of text, often a few sentences, that a retrieval system scores and an assistant may quote. Passages that make one complete claim are retrieved and quoted more cleanly than ones that depend on surrounding text.

Related: Answer-first content, Self-contained sentence

#Query fan-out

Rewriting one question into several narrower searches across subtopics and sources, then combining the results. Google describes AI Overviews and AI Mode as using it, which is why pages compete on sub-questions, not just the query as typed.

Related: AI Mode

#Question panel

A fixed, written set of questions real clients would ask, run against several AI engines on a schedule with every answer stored and dated. It turns AI visibility into something that can be measured and compared over time.

Related: Named share, Fresh session

#Retrieval-augmented generation (RAG)

A method in which an AI system first retrieves relevant documents, then writes its answer using them. Most assistants that answer questions about current businesses work this way.

Related: Grounding, Passage

#Robots.txt

A text file at a site's root telling crawlers which paths they may fetch, addressed to each crawler by its user-agent token, such as OAI-SearchBot, Claude-SearchBot or PerplexityBot. Go deeper →

Related: Crawl access, GPTBot

#Schema.org

A shared vocabulary for describing businesses, people, services and content in machine-readable form, used by Google, Microsoft and others. For law firms the relevant types are LegalService, Person and Service. Go deeper →

Related: Structured data

#Self-contained sentence

A sentence that still makes sense when quoted on its own, because it names its subject instead of relying on words like this, it or they to point back to earlier text.

Related: Passage, Answer-first content

#Server-side rendering

Producing a page's full HTML on the server, so its text is present before any script runs. Crawlers that do not execute JavaScript can only read pages built this way or pre-rendered.

Related: Crawl access

#Snippet controls

Directives such as nosnippet, data-nosnippet and max-snippet that limit what Google can show from a page, including in AI Overviews and AI Mode. Using them restricts visibility in those features.

Related: AI Overviews

#Structured data

Machine-readable markup, usually JSON-LD using the Schema.org vocabulary, that states facts about a page's subject. It helps search engines parse a business and can earn rich results; a May 2026 Ahrefs study found it did not measurably raise AI citations for pages already being cited. Google asks that it match the visible text. Go deeper →

Related: Schema.org, Knowledge graph

#Visibility sweep

GEO Expert's baseline measurement: forty questions a real client would ask, put to five answer engines from fresh sessions, with every answer and cited source recorded and dated, and three competitors measured on the same run. Go deeper →

Related: Question panel, Named share

What this means for you

See what the assistants say about you.

The free visibility sweep puts forty questions a real client would ask to five answer engines and records every answer and cited source, with three competitors on the same run.

Keep going