AI Search Glossary

The language of AI visibility.

Plain-language definitions for the terms, metrics and search features shaping generative AI discovery.

A

AI agent

AI systems

Definition. An AI system that may pursue a goal by planning and taking actions, often by using tools, browsing interfaces or completing a sequence of tasks. The degree of autonomy varies by system.

Use it when. The system does more than produce an answer and can act on a user's behalf.

AI answer

Foundations

Definition. A response generated or assembled by an AI-powered assistant, answer engine or search experience in response to a user's prompt.

Use it when. You need a platform-neutral term for the output being measured.

AI assistant

AI systems

Definition. A user-facing product that applies one or more AI models to answer questions, generate content, retrieve information or complete tasks.

Use it when. You mean the complete product experience rather than only its underlying language model.

AI Mode

Search features

Definition. Google Search's conversational AI search experience for exploring complex questions through generated responses, follow-up prompts and supporting links.

Use it when. You are specifically discussing Google's conversational AI search interface.

AI Overview

Search features

Definition. A generated summary that Google may show within search results, usually alongside links that help users explore the subject further.

Use it when. You are discussing Google's generated summaries, not AI answers in general.

AI recommendation

Measurement

Definition. A working measurement term for an AI answer that actively presents a brand, product or provider as a suitable option for the user's stated need.

Use it when. You need to distinguish an active suggestion from a passing mention.

AI search

Foundations

Definition. A broad term for search and discovery experiences that use AI to interpret questions, retrieve information and generate or organise answers.

Use it when. You want to cover both AI-native assistants and AI features within traditional search products.

AI SEO

Foundations

Definition. An informal industry umbrella term for applying search optimisation principles to discovery in AI-assisted and generative search experiences.

Use it when. Your audience understands SEO language and needs a bridge into AEO or GEO; clarify that AI SEO is an informal industry term.

AI share of voice (AI SOV)

Measurement

Definition. A brand's share of qualifying appearances—such as mentions or recommendations—compared with total qualifying appearances across the brand and a defined competitor set. The event type and denominator must be stated.

Use it when. You want to compare relative visibility across the same prompts, platforms and measurement period.

AI visibility

Foundations

Definition. The frequency and measured prominence with which a brand appears in AI-generated answers, recommendations and supporting citations for a defined prompt set. Prominence should be defined using measures such as recommendation position, citations or answer placement.

Use it when. You need an umbrella term for a brand's presence across multiple AI platforms, while recognising that visibility depends on the defined prompt set and measurement method.

Answer accuracy

Measurement

Definition. An assessment of whether an AI answer presents important facts about a brand, product or subject correctly and without material omissions.

Use it when. Visibility alone is insufficient and you also need to evaluate the quality of what is being said.

Answer engine

Foundations

Definition. A system designed to respond to questions with a direct answer or synthesis, often alongside links, rather than only presenting a list of links.

Use it when. You are describing the experience that AEO is intended to support.

Answer Engine Optimisation (AEO)

Foundations

Definition. An informal industry practice of making information clear, useful and authoritative enough to be selected for direct answers in search and AI-assisted experiences.

Use it when. You are discussing visibility inside answers; clarify that AEO is an industry term, not a guaranteed ranking method.

B

Baseline

Prompts & testing

Definition. The initial set of measurements captured before a campaign or change, used as the reference point for judging later movement.

Use it when. You need to show improvement or decline over 30, 60 or 90 days.

Benchmark

Prompts & testing

Definition. A reference result created from a consistent sample that helps compare performance across brands, industries, platforms or time periods.

Use it when. You are comparing a result with a broader peer group rather than only its own baseline.

Brand entity

Content & authority

Definition. A real-world organisation, product or brand that search and AI systems can connect with attributes, categories, people and external evidence.

Use it when. You are discussing how systems understand what a brand is and what it is associated with.

Brand mention

Measurement

Definition. Any appearance of a brand's name in an AI answer, whether or not the brand is recommended, described positively or cited as a source.

Use it when. You want the broadest measure of brand presence in answers.

Brand mention rate

Measurement

Definition. The percentage of valid tested answers in which a brand is named at least once.

Use it when. You need a consistent measure of presence across a controlled prompt sample.

Buyer prompt

Prompts & testing

Definition. A commercially relevant question a potential customer might ask an AI system while researching, comparing or selecting a product or provider.

Use it when. You are prioritising prompts connected with real purchase decisions rather than general curiosity.

C

Canonical URL

Content & authority

Definition. The preferred address for a page when the same or very similar content can be reached through more than one URL. A canonical signal tells search engines which URL you prefer, but does not guarantee that they will select it.

Use it when. You need to consolidate duplicate versions and signal which page should be treated as primary.

Citable content

Content & authority

Definition. Content that offers clear, verifiable and useful information—such as original data, definitions, methods or expert analysis—that other sources can legitimately reference.

Use it when. You are creating assets intended to earn legitimate citations or support grounded answers.

Citation

Measurement

Definition. A visible link or source attribution within an AI answer that identifies where supporting information came from.

Use it when. You are measuring source visibility, which is different from a brand merely being named.

Citation rate

Measurement

Definition. The percentage of valid tested answers that link to or identify a brand's domain as a supporting source, using a clearly stated denominator.

Use it when. You need to quantify how often a domain is used as evidence.

Context window

AI systems

Definition. The maximum amount of tokenised information an AI model or system can process in one interaction. Exact limits and how input and output are counted vary by model.

Use it when. You are explaining why long conversations or large documents may be truncated, summarised or handled differently.

Crawlability

Content & authority

Definition. The ability of an authorised crawler to access a webpage and its important textual content without being blocked by technical controls.

Use it when. You are diagnosing whether search systems can reach a page at all.

D

Digital PR for GEO

Content & authority

Definition. An informal industry label for earning legitimate editorial coverage, expert mentions and citations that strengthen the external evidence associated with a brand or subject.

Use it when. You are connecting off-site authority building with generative search visibility.

E

E-E-A-T

· Experience, Expertise, Authoritativeness and Trustworthiness

Content & authority

Definition. Google's framework for thinking about signals of experience, expertise, authoritativeness and trustworthiness. It is not a single standalone ranking factor or a checklist.

Use it when. You are assessing whether content demonstrates credible authorship, evidence and trust.

Embedding

AI systems

Definition. A numerical representation of content that helps systems compare meaning and retrieve semantically related information.

Use it when. You are explaining semantic retrieval rather than traditional exact-keyword matching.

Entity SEO

Content & authority

Definition. SEO work that helps search systems understand a real-world entity and its relationships, attributes and category associations.

Use it when. You are strengthening brand understanding beyond matching a page to individual keywords.

F

Featured snippet

Search features

Definition. A prominent extract from a webpage that Google may display in search results to answer a query quickly.

Use it when. You mean the established search-result feature, not a generated AI Overview.

Fine-tuning

AI systems

Definition. Further training a model on selected examples so its behaviour is better adapted to a task, style or domain.

Use it when. You mean changing model behaviour through additional training, not adding current facts through retrieval.

Foundation model

AI systems

Definition. A broadly capable model trained on large and varied datasets that can be adapted or applied to many downstream tasks.

Use it when. You are discussing the general-purpose model beneath a range of AI products and applications.

G

Generative AI (GenAI)

AI systems

Definition. AI that generates, transforms or synthesises content—such as text, images, audio, video or code—in response to an input.

Use it when. You need the broad technology category rather than the narrower term LLM.

Generative Engine Optimisation (GEO)

Foundations

Definition. An informal industry term for work intended to improve the likelihood that a brand, product or source is accurately represented in responses produced by generative AI systems.

Use it when. You are discussing brand visibility across conversational and generative search experiences; it is not a Google-defined ranking discipline.

Generative search

Foundations

Definition. A search experience that uses generative AI to synthesise an answer, often using retrieved sources and offering routes for further exploration.

Use it when. You are describing generated search responses rather than all uses of conversational AI.

Grounding

AI systems

Definition. The process of connecting an AI response to retrieved evidence or trusted source material so the answer is better supported.

Use it when. You are discussing the evidence used to support an AI-generated response.

H

Hallucination

AI systems

Definition. An inaccurate or fabricated AI-generated statement that can sound plausible despite not being properly supported.

Use it when. You are describing factual unreliability in generated output. "Confabulation" is a related term, but not always an exact synonym.

I

Indexability

Content & authority

Definition. A page's eligibility to be stored and considered in a search engine's index after it has been discovered and crawled.

Use it when. You are diagnosing whether an accessible page can be included in search results.

Internal link

Content & authority

Definition. A link from one page to another page on the same website, helping users and crawlers discover related content and understand site structure.

Use it when. You are connecting glossary definitions, articles, methodology and product pages.

K

Knowledge graph

Content & authority

Definition. A structured network of entities and the relationships between them, used to organise facts and support understanding beyond individual keywords.

Use it when. You are explaining how brands, people, places and categories can be connected as entities.

L

Large Language Model (LLM)

AI systems

Definition. A machine-learning model trained on large volumes of language data to interpret prompts and generate or transform text.

Use it when. You are referring to the underlying model, not the complete search product or assistant interface.

LLM optimisation

· Large language model optimisation

Foundations

Definition. An informal industry term for improving how content, entities or information are understood, retrieved or represented by LLM-powered experiences.

Use it when. Your audience uses LLM terminology; clarify whether you mean retrieval, content visibility or model engineering.

LLM ranking

Foundations

Definition. An informal phrase for a brand's position or prominence in AI answers. Unlike a traditional search ranking, it is not one fixed or universal position.

Use it when. Use cautiously; for measurement, prefer recommendation position, mention rate or AI share of voice.

llms.txt

Content & authority

Definition. A proposed voluntary text or Markdown file for presenting selected website information to some LLM-based tools. Google Search does not use the file, and other services may or may not.

Use it when. You are discussing experimental machine-readable publishing conventions—not a guaranteed ranking method.

Location sensitivity

Prompts & testing

Definition. The degree to which an AI answer or recommendation changes when the user's stated or inferred geographic location changes.

Use it when. You are testing local businesses, regulated markets, franchises or location-dependent availability.

M

Model version

Prompts & testing

Definition. The specific release or configuration of an AI model used during a test, which may affect the answer produced.

Use it when. You need enough testing context for results to be interpreted or reproduced responsibly.

Multimodal model

AI systems

Definition. An AI model that can work with more than one type of input or output, such as text, images, audio or video.

Use it when. You are discussing discovery experiences that interpret or generate more than text.

O

Organic search

Search features

Definition. Unpaid visibility within search results, distinct from advertisements or sponsored placements.

Use it when. You are distinguishing earned search discovery from paid media.

P

Personalisation

Prompts & testing

Definition. Changes to an answer influenced by information associated with a user, account, conversation history, preferences or prior activity.

Use it when. You are explaining why two people may receive different answers to similar prompts.

Platform coverage

Prompts & testing

Definition. The range of AI platforms included in a measurement programme. Report it separately from the proportion of platforms or prompts on which a brand appears.

Use it when. You want to distinguish the scope of testing from the visibility achieved within that scope.

Prompt

Prompts & testing

Definition. The instruction, question or context given to an AI system to shape the response it produces.

Use it when. You are naming the input submitted to an AI assistant or model.

Prompt coverage

Prompts & testing

Definition. The proportion of a defined prompt set for which a brand achieves the selected visibility condition, such as a mention or recommendation.

Use it when. You want to know how widely visibility is distributed across important buyer questions.

Prompt set

Prompts & testing

Definition. A controlled group of prompts selected for repeated testing around a brand, market, industry or stage of the buying journey.

Use it when. You need a consistent measurement sample rather than isolated screenshots.

Prompt tracking

Prompts & testing

Definition. Repeatedly testing defined prompts and recording changes in mentions, recommendations, citations and competitors over time.

Use it when. You are monitoring movement rather than conducting a one-off AI visibility audit.

Prompt variation

Prompts & testing

Definition. A deliberate change to the wording, context, location or constraints of a prompt used to test how sensitive an answer is to the input.

Use it when. You are evaluating whether visibility is robust or dependent on one exact phrase.

Q

Query fan-out

AI systems

Definition. A process in which a system generates several related searches to gather the different information needed to answer a broader question.

Use it when. You are explaining why one conversational query may create multiple retrieval opportunities.

R

Recommendation position

Measurement

Definition. The order in which a brand appears among the options actively recommended in an AI answer.

Use it when. You need a position-based measure without implying a permanent universal AI ranking.

Recommendation rate

Measurement

Definition. The percentage of valid tested answers that actively present a brand as a suitable option for the user's stated need.

Use it when. You need to quantify active recommendations separately from mentions.

Result volatility

· Answer volatility

Prompts & testing

Definition. The degree to which answers, sources or recommendations change across repeated tests, platforms, model versions or measurement periods.

Use it when. You are explaining why a single answer is not enough to establish reliable visibility.

Retrieval-Augmented Generation (RAG)

AI systems

Definition. A method that retrieves relevant information from external or connected sources and uses it as context when generating a response.

Use it when. You are explaining how an AI answer can be grounded in current or selected source material.

Rich result

Search features

Definition. A search result enhanced with additional visual or interactive information made possible by supported structured data and other signals.

Use it when. You mean a supported search appearance, not an AI recommendation or citation.

robots.txt

Content & authority

Definition. A site-level file containing requests that specified web crawlers follow about which paths they may or may not crawl. It guides compliant crawlers but is not a security barrier or a reliable way to remove an already indexed page.

Use it when. You are managing crawler access or diagnosing crawl behaviour.

S

Schema markup

Content & authority

Definition. Structured data on a webpage that uses the Schema.org vocabulary to describe visible content in a machine-readable form. It is commonly implemented with JSON-LD, Microdata or RDFa.

Use it when. You are implementing structured data. Do not present it as a special or guaranteed AI-ranking tactic.

Search engine index

Search features

Definition. A search engine's organised store of discovered and processed web content that may be considered for retrieval and ranking.

Use it when. You are distinguishing being known to a search engine from merely existing on the web.

Search Engine Optimisation (SEO)

Foundations

Definition. The practice of improving a website so search engines can discover, understand and surface its pages for relevant searches.

Use it when. You are discussing organic search discovery, crawlability, relevance and website rankings.

Sentiment

Measurement

Definition. The overall positive, neutral or negative framing of a brand within an AI answer, assessed using a documented classification method.

Use it when. You need to evaluate how a visible brand is portrayed, not simply whether it appears.

Source authority

Content & authority

Definition. The perceived credibility and subject relevance of a source based on signals such as expertise, reputation, evidence and independent recognition.

Use it when. You are assessing why one source may be trusted or cited over another.

Structured data

Content & authority

Definition. Information presented in a standard machine-readable format to describe a page's content and the entities it contains.

Use it when. You mean the broader technical concept; schema markup is one way of implementing it.

T

Third-party corroboration

Content & authority

Definition. Independent external evidence that supports claims about a brand, such as credible reviews, editorial coverage, directories, credentials or customer references.

Use it when. You are strengthening claims with evidence beyond the brand's own website.

Token

AI systems

Definition. A unit into which text or other input is divided for processing by an AI model; a token may be a word, part of a word or punctuation.

Use it when. You are discussing model input limits, output length or usage-based processing.

Top-three recommendation rate

Measurement

Definition. The percentage of valid tested answers in which a brand is actively recommended within the first three eligible options, where at least three options are listed.

Use it when. You need a stricter prominence measure than general recommendation rate, using a defined first-three rule.

U

Unlinked brand mention

Content & authority

Definition. A reference to a brand on an external page that does not include a clickable link to the brand's website.

Use it when. You are measuring brand recognition separately from backlink acquisition.

V

Valid answer

Prompts & testing

Definition. A completed response that meets the documented criteria for inclusion in a measurement sample and is not excluded because of an error or failed test.

Use it when. You need to define the denominator used in visibility calculations.

Visibility score

Measurement

Definition. A combined indicator built from selected visibility metrics. Its meaning depends on the provider's formula, weighting and prompt sample.

Use it when. You are summarising performance; publish the methodology and avoid comparing scores from unrelated tools as if they were equivalent.

Z

Zero-click search

Search features

Definition. A search journey in which the user gets the information they need from the results experience without visiting another website.

Use it when. You are explaining why visibility and brand influence can matter even when they do not produce an immediate website click.

Common questions

Can a website rank in ChatGPT?
A brand or website can appear prominently for a defined set of prompts, but there is no single universal ChatGPT ranking. Measure repeated mentions, recommendations, positions and citations across controlled tests.
What is the difference between AEO and GEO?
AEO focuses on making information suitable for direct answers. GEO focuses more specifically on visibility and accurate representation in generative AI responses. Both are informal industry terms. In practice, both depend on strong content, technical accessibility and credible evidence.
Does schema markup improve AI visibility?
Structured data can help systems understand page content and may support eligible search features, but it is not a special AI-ranking switch and does not guarantee a mention or citation.
Is an llms.txt file required?
No. It is a proposed convention that some tools may use. Google Search does not use it for Search or its generative AI features, so it should not be treated as a Google ranking requirement.
How often do AI recommendations change?
They can change between runs and over time because of model updates, new source material, prompt wording, location and personalisation. Reliable measurement therefore uses a repeatable prompt set and a defined testing cadence.
How should AI visibility be measured?
Start with commercially relevant prompts, record valid answers across selected platforms, classify mentions, recommendations and citations consistently, and compare results over a stated period against a baseline and competitor set.