GEO glossary

Jump to: A · C · D · E · F · G · H · L · M · P · R · S · Z · Visigent terms

Industry terms

A

AI Overviews

Google's generated summary shown above or within search results, drawing on a set of cited pages. A distinct surface from Gemini, with its own citation behaviour.

AI visibility

How often, and how prominently, a brand appears in the answers AI systems give to relevant prompts. Measured, not ranked: there is no single position, only presence across many answers.

AI referral traffic

Visits that arrive from an AI engine (ChatGPT, Perplexity, Gemini, Claude, Copilot and others), identifiable by referrer or UTM in analytics. Typically small in volume and high in intent.

AI Mode

Google's conversational search surface, separate from AI Overviews, that answers multi-step queries with generated responses and citations.

Answer engine

A system that responds to a question with a synthesised answer rather than a list of links. ChatGPT, Perplexity, Gemini, Claude and Google's AI surfaces are answer engines; classic web search is not.

Authoritative list mention

A brand's inclusion on a third-party "best of", roundup, directory or comparison page. One of the strongest signals an engine uses when deciding whom to recommend.

C

Citation

An explicit reference by an AI answer to a source page or domain, usually shown as a link or footnote. A citation says the engine used the page; it does not mean the brand was recommended.

Crawler (AI)

The bots AI companies use to fetch web pages for training or retrieval, such as GPTBot, ClaudeBot and PerplexityBot. Sites can allow or block them in robots.txt; blocking removes the site from that engine's retrieval. The technical layer →

D

Description accuracy

Whether an AI system's account of a business — what it does, where it operates, what it offers — matches the facts. Errors here are usually caused by gaps or contradictions in what the business has published.

E

Entity

A thing the model treats as distinct and identifiable: a company, product, person, place. Engines answer about entities, not pages; a business the model cannot resolve as one entity is described inconsistently or left out.

Entity clarity

How unambiguously an AI system can identify a business and its attributes across sources. Consistent names, addresses, offerings and structured data raise it; contradictions lower it. The entity play →

F

Fan-out (query fan-out)

The way modern engines expand a single prompt into several sub-queries, retrieve for each, and synthesise. A brand can be pulled in through any sub-query, which is why coverage of adjacent topics matters.

G

Generative Engine Optimization (GEO)

The practice of increasing a brand's presence, accuracy and prominence in the answers AI systems generate. It covers on-site content and structure, off-site mentions and reviews, and measurement of the answers themselves. What is GEO? →

Grounding

The retrieval step in which an engine fetches current web content to base its answer on, rather than relying only on training data. Grounded answers cite sources; ungrounded ones reflect what the model already believes.

H

Hallucination (brand)

An AI answer that states something false about a business: a service it doesn't offer, a location it doesn't serve, a price it doesn't charge. Usually a retrieval gap being filled by inference. Field guide →

L

LLM (large language model)

The underlying model that generates text in an AI engine. Different products use different models; the same brand can be described differently by each.

M

Mention

A brand named in an AI answer, with or without a link. Mentions are the unit of share of voice; citations are the unit of sourcing. The mention is the signal →

P

Prompt

The question a user asks an engine. GEO measurement runs a defined set of prompts repeatedly to see whose brands appear. Optimize for the prompt →

R

Recommendation

An answer in which the engine names a brand as a suggested choice for the user's need. Stronger than a mention; the outcome GEO aims at.

Recommendation rate

The share of relevant prompts, over repeated runs, in which a brand is recommended.

Retrieval-augmented generation (RAG)

The architecture in which a model retrieves documents and generates an answer from them. Most grounded AI search works this way; it is why current, readable pages matter.

S

Share of voice (AI)

A brand's share of mentions across a set of prompts, relative to competitors. Tracked over time and by engine.

Structured data (schema)

Machine-readable markup (schema.org) that states facts about a page's subject: organisation, product, FAQ, article. Reduces the guessing an engine has to do. Structured proof →

Z

Zero-click

A search or AI interaction that ends without a visit to any website because the answer was given in place. In answer engines it is the default, which is why being named matters more than being linked. The zero-click era →

Visigent terms

Terms Visigent uses for its own methodology and product. Each links to the page that defines it fully.

GEO Score

A 0–100 score of how eligible a business is to be recommended by AI systems, computed weekly across four signals. Full definition →

Beacon

Visigent's model of a business: everything it knows about itself, mapped against every page it has published, with a decision for each — keep, update, consolidate, build. What Beacon is →

Signal Hierarchy

Visigent's published ranking of the sources AI systems weight when recommending a brand, from authoritative list mentions down. The Signal Hierarchy →

Managed program

Visigent's service in which its team measures, models, plans, executes and proves a customer's GEO under the customer's approval. The program →

The five steps of Visigent's method. Product overview →

Understood

Visigent's word for the state in which an AI system has a complete, accurate, consistent record of a business and nothing left to infer. The precondition for being recommended. What understood looks like →

Written by Tom Howell, Co-Founder & CEO. Reviewed by Product and Editorial. Updated September 2026.