New: automated hallucination alerts are live. Learn more →

LLM Metrix logoLLM Metrix

Glossary of Terms

147 definitions covering AEO, GEO, AI visibility, and the terminology behind AI-powered search optimization.

ABCDEFGHIJKLMNOPQRSTUVWXYZ

A

AEO

Answer Engine Optimization — The practice of optimizing content to appear prominently in AI-generated answers and responses.

Answer Engine

An AI system that generates direct answers to user queries rather than returning a list of links. Examples: ChatGPT, Claude, Perplexity, Gemini.

AI Overviews

Google's AI-generated summaries placed above traditional search results, powered by Gemini with live web retrieval — one of the highest-value AEO placement opportunities.

AI Agent

An AI system that autonomously plans and executes multi-step research tasks — browsing pages, following links, and synthesizing findings — rather than answering a single query. Changes the rules for brand visibility as engines like ChatGPT and Perplexity adopt agentic modes.

Anchor Text

The clickable text of a hyperlink — when other sites link to your pages with category-relevant anchor text, it signals to AI retrieval systems what your brand and pages are about, influencing category association and retrieval ranking.

AI Recommendation

A suggestion made by an AI engine for a product, service, or tool in response to a user query — the AI equivalent of a trusted referral, delivered at scale. Higher purchase-intent than most advertising because the user is actively asking for a recommendation in a defined context.

Author Authority

The recognized expertise and credibility of the individual who wrote a piece of content — a key E-E-A-T signal. AI retrieval systems score documents partly on author credibility, making named bylines, author bio pages, and Person schema markup directly actionable for improving citation rates.

Answer Box

A formatted display in Google SERPs that presents a direct answer — extracted from a third-party page — at the top of results. The behavioral and structural precursor to AI-generated answers. Pages that consistently win answer boxes typically have the direct-answer structure that AI engines also prefer for citations.

AI Mode

Google's dedicated AI-powered conversational search interface — a full-page multi-turn research experience powered by Gemini, separate from standard Google Search. Announced at Google I/O 2025. Higher value for considered-purchase and B2B brands because users conduct deeper research in AI Mode than in standard AI Overviews.

Agentic AI

AI systems that can take multi-step actions toward a goal — planning, calling tools, browsing, and executing tasks — rather than just answering a single prompt. Agentic shopping and research assistants increasingly mediate brand discovery, making accurate, machine-readable brand information more important.

AI Shopping Assistant

A generative AI assistant that helps users research and choose products conversationally — for example Amazon Rufus. AI shopping assistants are grounded largely in commerce data (listings, attributes, reviews), making accurate, complete product information a direct visibility lever.

AI Referral Traffic

Website visits that come from users clicking a citation or link in an AI engine's answer (for example from Perplexity, Google AI Overviews, or ChatGPT). It's the most direct owned-data signal of AEO working — but it captures only click-throughs, undercounting the larger zero-click impact of being seen in answers.

B

Brand Sentiment

The overall tone (positive, neutral, negative) with which AI engines discuss your brand in their responses.

Brand Recall

The strength with which AI models associate your brand with a particular category or use case — determines how often your brand surfaces in category-level queries without being explicitly named.

Brand Safety

The practice of detecting and correcting harmful, inaccurate, or misleading AI-generated representations of your brand — before they reach customers at scale across every engine.

Brand Entity Disambiguation

The process by which AI systems resolve which specific entity a brand name refers to — especially important for brands with common words in their names. Strong disambiguation signals (consistent category anchoring, Wikidata entries, schema markup) reduce hallucination and category confusion in AI responses.

Brand Mention

Any appearance of your brand name in an AI-generated response, whether or not a URL link is included. The broader category containing citations as a subset — all citations are brand mentions, but not all brand mentions are citations.

Brand Salience

The degree to which a brand comes to mind in a relevant buying situation. In the AI era, brand salience is increasingly mediated by AI engines — brands absent from AI category recommendations never enter the buyer's mental consideration set, making AI visibility a new dimension of salience strategy.

Brand Equity

The commercial value a brand name adds beyond its product's functional utility. In AI visibility: high-equity brands have richer training data signals and receive more confident AI recommendations. Brand equity also creates a durable competitive moat — displacing a well-known brand from AI consideration sets requires generating far more training data volume.

BM25

A classical document retrieval algorithm that ranks documents by term frequency and inverse document frequency — used in traditional search engines and as a first-pass retrieval stage in many AI search pipelines. The technical grounding for why vocabulary matching and using audience-native terminology still matters in AI-indexed content.

Brand Voice

The consistent personality, tone, and terminology with which a brand communicates. Relevant to AI visibility because consistent brand voice creates coherent entity associations in AI training data — and content written with authoritative, direct voice is more strongly associated with confident AI recommendations than hedging or vague language.

Backlink Profile

The complete collection of external links pointing to a website — including quantity, quality, domain diversity, anchor text, and topical relevance. A strong backlink profile is a primary input into domain authority scores that AI retrieval systems use as trust signals when selecting content to cite.

Benchmark

A standardized test used to measure and compare AI model capabilities (e.g. reasoning, coding, factual accuracy). Benchmarks shape which models engines deploy and how they behave — and benchmark-driven model updates can shift how your brand is represented over time.

Brand Authority

The degree to which a brand is recognized as a credible, expert source in its category — built through reputation, expertise, and corroboration across the web. AI engines lean on authority signals when deciding which brands to name and trust, making brand authority a core driver of AI visibility.

C

Citation

A reference to an external source that an AI engine includes in its response, either as an inline link, footnote, or attribution.

Content Gap

A topic or query where competitors are cited by AI engines but your brand is not — identifying and closing content gaps is one of the highest-leverage AEO strategies.

Competitive Displacement

When a competitor's brand appears in an AI response in place of yours — the AI-era equivalent of losing search rankings, with compounding effects across all instances of a query.

Content Freshness

The recency of web content as a ranking signal for AI engines — particularly RAG-powered systems that prefer recently updated pages when retrieving sources for citation.

Context Window

The maximum amount of text an LLM can process in one interaction — including query, retrieved documents, and system instructions. Determines how much of your content is actually read and cited in a single AI response.

Chunking

The process of splitting long documents into smaller segments for RAG indexing — chunks are what retrieval systems actually retrieve, not whole pages. Heading structure and section clarity determine where chunk boundaries fall and which content gets cited.

Citation Velocity

The rate at which your brand gains new AI citations over time — how quickly you're being added to AI-generated responses. A rising velocity signals an effective AEO strategy; flat or declining velocity signals a plateau even if total citation count is high.

Content Velocity

The rate at which a brand publishes new, indexable content over time. High content velocity within a focused topic cluster builds topical authority faster and increases AI citation coverage — compounding over time as each new piece expands query surface area.

Crawl Budget

The number of pages a web crawler will fetch from your site within a given period. AI crawlers (GPTBot, PerplexityBot) have independent crawl budgets — pages not crawled are not eligible for AI citation. Wasted budget on duplicate or thin pages means important content may go unindexed.

Consideration Set

The shortlist of brands a buyer actively evaluates before making a purchase decision — typically 2–5 options. AI engines now form consideration sets directly: the brands named in an AI response to a category query become the buyer's shortlist, making AI mention presence prerequisite to entering evaluation.

Conversational Search

Querying an AI system using natural language sentences rather than keyword fragments — the dominant query mode for AI engines. Conversational queries contain context, constraints, and persona signals that change which brands get recommended. Content written to address specific scenarios and constraints performs better for conversational retrieval.

Citation Diversity

A measure of how many different source types and domains mention your brand — vs. concentrated coverage from few sources. High citation diversity signals broad market recognition to AI engines and provides resilience; concentrated coverage from few sources may not generalize across query types.

Canonical URL

The preferred, definitive URL for a page — specified with a `<link rel='canonical'>` tag. Consolidates crawl budget and link equity to a single URL when duplicate or near-duplicate versions exist. Missing or incorrect canonical tags fragment authority across URL variants, reducing AI retrieval priority.

Content Hub

A dedicated site section aggregating related content — articles, guides, glossary entries, case studies — organized around a central topic. Creates topical authority signals at scale through content breadth and internal link density. More AI-visibility-effective than a chronological blog because all content remains equally discoverable.

Category Leadership

The position of being recognized as the foremost brand in a product or service category. In AI responses: category leaders are named first in broad category queries and used as benchmarks against which other brands are compared. Challenger brands can target category leadership in specific niches before competing for broad category ownership.

Content Audit

A systematic review of all published content on a website — assessing quality, accuracy, relevance, and performance — to determine what to keep, update, consolidate, or remove. Essential for AEO because AI crawlers index your entire content footprint; outdated or thin content wastes crawl budget and creates brand safety risks.

Cosine Similarity

A metric measuring how similar two vectors are — used in AI search to compare query embeddings to document embeddings. Documents with higher cosine similarity to the query are retrieved first. Explains why semantically precise, vocabulary-consistent content outperforms keyword-dense content in AI retrieval.

Citation Rate

The share of relevant AI answers in which your content is cited as a source. A core AEO metric for citation-first engines: it measures not just whether your brand is mentioned, but whether your pages are credited as the source the answer drew from.

Co-Citation

When two brands or sources are mentioned together across many pages, signaling to engines that they're related or comparable. Co-citation helps AI systems build associations — being consistently mentioned alongside category leaders can position your brand within that set.

Conversational AI

AI systems designed to interact through natural, multi-turn dialogue — the category that includes assistants like ChatGPT, Gemini, and Claude. Conversational AI is the interface where much AEO now plays out, since users refine questions across turns rather than issuing one keyword query.

Crawlability

How easily automated crawlers — including AI crawlers — can access and read your pages. If content is blocked, hidden behind scripts, or slow to load, retrieval-based engines may never see it, making crawlability a prerequisite for AI visibility.

D

Domain Authority

A metric of your website's overall trustworthiness and credibility based on backlink profile, age, and content quality — influences AI visibility.

Digital PR

The practice of earning online press coverage, editorial mentions, and backlinks through story-driven outreach. One of the highest-leverage AEO activities because it produces third-party citations on high-authority domains — the kind of signal AI engines weight most heavily for brand authority.

Duplicate Content

Substantial content blocks appearing at multiple URLs — within a site or across domains. Fragments crawl budget and dilutes link equity by splitting authority across URL variants. Fixed by canonical tags, 301 redirects, and page consolidation. Common sources: URL parameters, HTTP/HTTPS variants, and syndicated content without canonical tags.

Dwell Time

The amount of time a user spends on a page after arriving. A content quality signal that correlates with AI citation-friendly attributes: comprehensive coverage, direct answers, and structured depth. Content optimized for human engagement quality tends to also meet AI retrieval quality standards.

Deep Research

An AI capability (offered by several engines) that autonomously runs many searches, reads numerous sources, and synthesizes a long, cited report on a topic. Deep-research modes read far more sources than a normal query, rewarding brands with broad, authoritative, well-structured coverage.

Digital Shelf

The collective online spaces where products are discovered and evaluated — marketplaces, retailer sites, and now AI shopping assistants. As AI mediates product discovery, the digital shelf extends into AI answers, where accurate product data and reviews determine whether you're recommended.

E

E-A-T

Expertise, Authoritativeness, Trustworthiness — Google's quality framework that also strongly influences how AI systems evaluate and cite content.

Entity

A uniquely identifiable concept — brand, person, product, or place — recognized by AI systems as a distinct object with known attributes and category relationships.

Embeddings

Mathematical vector representations of text that encode meaning rather than exact words — the core technology behind semantic search and RAG retrieval, enabling AI engines to find conceptually relevant content regardless of keyword overlap.

E-E-A-T

Experience, Expertise, Authoritativeness, Trustworthiness — Google's content quality framework. RAG-powered AI engines, especially Google AI Overviews, use similar signals to decide which sources to cite. Author attribution, original research, external citations, and factual accuracy all contribute.

Evergreen Content

Content designed to remain accurate and valuable over an extended period without frequent rewrites. The foundation of AI visibility strategy: evergreen pages accumulate citation signals, stay in RAG indexes, and generate stable brand-concept associations in LLM training data — compounding in value over months and years.

Extractability

How easily an AI engine can lift a clear, self-contained answer out of your content to quote in its response. High-extractability content leads with direct answers, uses statement headings, and presents facts in lists, tables, and FAQs — making it easy to retrieve and cite.

I

Impression Rate

The percentage of tracked queries for which your brand appears in AI-generated responses — the broadest measure of AI brand presence.

Indexability

The degree to which AI engines and search systems can discover and crawl your content — a prerequisite for appearing in RAG-powered AI responses.

Inference

The real-time process by which a trained AI model generates a response to a user's prompt. Distinct from training (which happens once in advance). Brand visibility is determined at inference time by training data recall, live retrieval, and system prompt configuration.

Internal Linking

Linking from one page on your website to another page on the same domain. Distributes link equity, guides AI crawlers to important content, and signals topical relationships between pages. Well-structured internal linking is one of the most controllable factors in improving AI retrieval coverage for strategically important content.

In-Context Learning

An LLM's ability to adapt to new tasks or incorporate new information from content provided within the current session — without retraining. Explains why well-structured, dense content improves citation quality: when retrieved by RAG, it provides richer in-context information for more accurate and specific AI responses.

Information Gain

How much new, unique value a page adds beyond what's already widely available. Content with high information gain — original data, fresh analysis, first-hand experience — is more likely to be cited by AI engines, which favor sources that contribute something not found everywhere else.

L

Listed Mention

When your brand appears in a list alongside competitors, typically with less emphasis than a first or prominent mention.

LLM

Large Language Model — The foundational AI technology behind chatbots like ChatGPT, Claude, and Gemini that powers natural language understanding and generation.

Lift

The measurable improvement in AI visibility metrics resulting from a specific content action or optimization — used both as a forecast (projected lift) before acting and as an attribution metric after shipping.

LLMO

Large Language Model Optimization — an emerging industry term for improving brand visibility in LLM outputs, used interchangeably with AEO and GEO by different practitioners and publications.

Link Equity

The authority value that flows from one page to another through hyperlinks — the higher the linking page's authority, the more equity passes. For AI visibility: link equity from relevant, authoritative sources improves retrieval ranking for RAG engines; the semantic context of linking text also shapes brand-category associations.

llms.txt

A proposed standard — a Markdown file at the root of your domain (yourdomain.com/llms.txt) that gives AI systems a curated, prioritized map of your most important content. A counterpart to robots.txt and sitemap.xml, but written for large language models rather than search crawlers.

Long-Tail Query

A specific, lower-volume search phrase, often phrased as a detailed natural-language question. Long-tail queries are where AEO often wins first: they're less contested, map cleanly to focused content, and match how people actually talk to AI assistants.

M

Mention Positioning

The location and prominence of your brand within an AI response — can be first mention (most valuable), prominent mention (medium), or listed mention (lower value).

Model Alignment

Training that shapes an LLM to be helpful, harmless, and honest — using techniques like RLHF. Alignment policies govern how models handle brand recommendations and can suppress or qualify mentions in sensitive categories, independent of retrieval quality.

Multimodal

AI models that process multiple data types — text, images, charts, video, and code. As AI search becomes multimodal, product screenshots, branded infographics, and video transcripts become part of your citable content surface.

Multi-Turn Conversation

An AI interaction spanning multiple exchanges where context accumulates across messages. Follow-up queries in a session are often higher-intent and more specific than opening queries — brands with content addressing refined, contextual queries appear in these higher-value moments.

Model Distillation

A technique for creating smaller, faster AI models by training them to replicate a larger model's behavior. Consumer AI products often run distilled models that may have less brand-specific knowledge than full-size models — especially for niche or newer brands whose training data signals were compressed out in the distillation process.

Model Context Protocol (MCP)

An open standard for connecting AI assistants to external tools and data sources through a common interface. MCP lets models retrieve live information and take actions via standardized 'servers,' expanding how and where AI systems access brand and product data.

Model Card

A document published by a model's maker describing its capabilities, training, intended uses, and limitations. Model cards help you understand a model's knowledge cutoff and behavior — context that matters when diagnosing why an engine represents your brand a certain way.

P

Prominent Mention

When your brand is featured in a key position (introduction, recommendation, or highlighted section) of an AI response.

Prompt

A natural language input submitted to an AI engine by a user — understanding how prompts are phrased and which brands they surface is foundational to AEO research.

Position Drift

The gradual or sudden shift in where your brand appears within AI responses over time — moving from first mention to listed mention, or from prominent to buried — a key early warning signal for eroding AI visibility.

Passage Indexing

The ability of retrieval systems to index and rank individual passages within a page independently — meaning a single highly relevant section can be retrieved and cited even if the overall page isn't a top result.

Prompt Engineering

The practice of designing and refining text inputs to AI systems to produce more accurate or targeted outputs. In AEO, relevant for designing effective monitoring queries that approximate real buyer behavior, and for understanding how system prompts shape AI product behavior.

Pillar Page

A comprehensive, long-form content piece that covers a broad topic area in full, serving as the authority hub for a topic cluster with links to all supporting cluster pages. Retrieved frequently by RAG systems for a wide range of related queries due to its broad semantic coverage.

PerplexityBot

The web crawler operated by Perplexity AI that indexes content for its RAG-first retrieval system. Perplexity displays inline citation links prominently, making PerplexityBot citations particularly traffic-visible. Allowing PerplexityBot access in robots.txt is recommended for any brand pursuing AI citation presence.

Page Authority

A metric predicting how well a specific page will rank, based on the quantity and quality of links pointing to that individual URL — distinct from domain authority. For AI retrieval, pages with higher page authority are retrieved more frequently. Direct link-building to strategic content pages improves both PA and AI retrieval priority.

Prompt Injection

A security issue where malicious instructions hidden in content (or user input) manipulate an AI system into ignoring its original instructions. Relevant to brand safety because untrusted web content an AI retrieves could attempt to alter how a brand is represented.

People Also Ask

The expandable list of related questions Google shows in search results. People Also Ask is a useful map of how real users phrase questions in your space — making it a practical source for the question-style content that AEO and FAQ optimization reward.

S

SERP

Search Engine Results Page — Traditional Google search results. Different from AI responses but related to overall online visibility strategy.

Share of Voice

The percentage of AI responses that mention your brand compared to competitors for a given set of queries or topics.

Semantic Search

Search methodology that interprets the meaning and intent behind a query rather than matching exact keywords — the foundation of how all modern AI engines process user queries.

Structured Data

Machine-readable Schema.org markup added to web pages that explicitly declares entity attributes and relationships — a direct technical lever for how AI systems represent your brand.

System Prompt

Hidden instructions given to an LLM before any user interaction that configure its behavior, citation preferences, and content policies — the primary reason the same model behaves differently across ChatGPT, Perplexity, Claude, and Copilot.

Semantic Triple

A structured factual statement consisting of subject, predicate, and object — the fundamental unit of knowledge graphs. Example: '[Brand] → is a type of → [category]'. AI engines learn brand attributes through these relationships, encoded in schema markup, Wikidata, and factual text.

Share of Search

The percentage of all branded searches in a category that include your brand name — a leading indicator of market share. Complements AI share of voice: share of search captures navigational intent from buyers who know you; AI share of voice captures discovery intent from buyers who don't yet.

Source Authority

The credibility score AI retrieval systems assign to a domain or page when selecting content to cite — determined by backlink profile, publishing history, author attribution, content quality, and topical consistency. The single most impactful structural factor in RAG-based citation selection.

SGE (Search Generative Experience)

Google's public testing name for what became AI Overviews — AI-generated answer summaries in Google Search. Tested in Search Labs 2023–2024 before full launch as AI Overviews at Google I/O 2024. Still widely used in the SEO/AEO industry as shorthand for AI-generated answers in Google Search.

Search Intent

The underlying goal behind a user's search query — informational, navigational, commercial investigation, or transactional. Synonymous with query intent. AI engines match content to intent before retrieving; a mismatch between content type and query intent prevents citation regardless of content quality.

Share of Model

A metric for how often your brand appears in AI answers for a category relative to competitors — the AI-era analog of share of voice. It reflects your slice of the 'mindshare' an AI engine expresses when asked about your space, across many queries and engines.

Source Attribution

How an AI engine credits the sources behind its answer — via inline citations, links, or a sources list. Strong source attribution makes AEO measurable, because you can see which pages an engine relied on to construct its response.

Synthetic Data

Artificially generated data used to train or fine-tune AI models, rather than data collected from the real world. As models increasingly train on synthetic and AI-generated content, the authority and consistency of original, credible brand information becomes more important.

Semantic HTML

HTML that uses meaningful elements (headings, lists, tables, articles) to convey the structure and role of content. Semantic HTML makes pages easier for crawlers and AI systems to parse and extract from — a quiet but real technical lever for AEO.

T

Topic Authority

Your recognized expertise in a specific domain or subject area — established through comprehensive content, expert authorship, and consistent publishing.

Training Data

The corpus of text and information that AI models learn from. Your content in training data influences how AI systems represent your brand.

Topical Depth

The comprehensiveness of content coverage on a given subject — AI engines strongly favor sources with deep topical coverage when generating and citing responses.

Token

The basic unit of text LLMs process — roughly 0.75 words each. Context windows, retrieval budgets, and API costs are all measured in tokens; understanding tokens explains why only portions of long pages get read by AI engines.

Temperature

An LLM parameter controlling output randomness — low temperature produces consistent, predictable responses; high temperature produces varied ones. The primary reason the same query returns different brand mentions across monitoring runs.

Topic Cluster

A content architecture where a central pillar page covers a broad topic, surrounded by supporting cluster pages covering related subtopics — all interlinked. Signals topical authority to AI retrieval systems by demonstrating comprehensive coverage of a subject area.

Thought Leadership

Content that shapes how an industry thinks about a topic — establishing the author or brand as an original authority rather than just a practitioner. Produces outsized AI citation value because it generates third-party citation chains, earns high-authority backlinks, and may shape the vocabulary AI engines use to describe a domain.

Topical Map

A structured plan mapping all topics, subtopics, and content pieces needed to establish full topical authority in a domain — defining pillar pages, cluster pages, and their interlinkage. Enables comprehensive query cluster coverage that builds the topical authority signals AI retrieval systems use to score domain expertise.

Thin Content

Web pages with insufficient depth, originality, or factual density to merit strong AI citation — high word count without specific citable claims, restated information without original perspective, or superficial coverage of too-broad topics. Wastes crawl budget, dilutes topical authority, and increases hallucination risk when retrieved.

Tokenization

The process of breaking text into smaller units (tokens) that a language model processes — roughly words or word-pieces. Tokenization underlies how models read content and how limits like context windows are measured, since models think in tokens rather than characters or words.

Back to Knowledge Base