Plain English definitions for the terms that matter in AI search, from how answer engines pick their sources to how brands measure whether the machines are getting them right.
Agentic Search
Search that an AI agent performs on behalf of a user, researching vendors, comparing prices, and even completing purchases. Increasingly, the reader of your website is a piece of software rather than a person.
AGI (Artificial General Intelligence)
A future form of AI that could match human ability across nearly all intellectual work, which does not exist yet but drives much of the industry’s ambition.
AI Agent
An AI system that takes action instead of simply answering questions, handling multi-step tasks like research, comparison, and purchasing with little supervision.
AI Crawlers
The bots that answer engines send out to read the web, including GPTBot, ClaudeBot, PerplexityBot, and Google-Extended. If these bots cannot reach or render your content, their models never learn that it exists.
AI Mode
Google’s conversational AI search experience, which works like a chat where follow-up questions replace new searches and cited sources replace the traditional ten blue links.
AI Overviews
The AI-generated summary that Google places above traditional search results, assembled from multiple sources. For many queries it is the only thing a searcher reads.
AI Referral Traffic
Website visits that arrive from AI platforms. These visitors are often close to a purchase because the AI answer educated them before they ever clicked.
AI Search
Search that happens inside an AI assistant or answer engine, where the system reads many sources and writes a single conversational answer instead of returning a list of links. The system decides for the user which brands get named and which get left out.
AI Visibility
Whether and how often your brand appears in AI-generated answers to the prompts that matter to your business. Visibility is the presence question, asking whether you are in the answer at all. Often represented as the number of times you appear in a set of answers (e.g. 99 appearances in 100 responses = 99% visibility).
Answer Engine
Any AI system that responds to a query with a written answer rather than a page of ranked results. ChatGPT, Perplexity, Gemini, Copilot, Claude, and Google’s AI features are examples.
Answer Engine Optimization (AEO)
A close cousin of GEO that focuses on structure, meaning content written to answer specific questions directly so an AI system can lift the answer cleanly. Many teams use GEO and AEO to mean the same thing.
Answer-First Formatting
Structuring content so the direct answer comes first and the supporting detail follows, because AI systems extract passages.
Answer Variability
The natural differences in AI responses when the same prompt runs more than once, caused by random sampling and constant model updates. Credible measurement of AI search relies on repeated sampling rather than single spot checks.
Artificial Intelligence (AI)
Software that performs tasks we associate with human intelligence, such as understanding language, recognizing patterns, and making decisions. It is the umbrella over everything else in this glossary.
Brand Fidelity
How accurately AI systems represent your brand, beyond whether they simply mention it. Brand fidelity asks whether the model places you in the right category, knows which products are current, describes your real customers, and repeats your actual differentiators. It also asks whether the model confuses you with a competitor or with an outdated version of your own company. Visibility measures presence, while brand fidelity measures accuracy. Term coined by Adam Malamis, founder of Gander.
Branded Prompt
A prompt that names your company or product, which tests what the models actually believe about you rather than whether they know you exist.
Buyer-Journey Prompts
The full sequence of questions a buyer asks an AI, from early education through final evaluation. Brands that appear in the early educational prompts often enter the comparison stage as the presumed favorite.
Citation
An AI answer that references or links to a specific page as a source, revealing which content the engines trust and which sites feed the answers about your category.
Citation Gap
Prompts and topics where AI engines cite your competitors’ content but not yours, which makes citation gaps the most direct to-do list that AI search offers a content team.
Citation Share
The percentage of all citations in a set of AI answers that point to your domain, serving as the AI search equivalent of ranking share.
Comparative Prompt
A prompt that asks an AI to weigh options against each other, such as one brand versus another. The answer works like an instant analyst report built from whatever content the AI has learned to trust.
Content Fidelity
How faithfully AI answers reproduce the substance of your content when they draw on it. High content fidelity means your facts, figures, claims, and positioning survive the trip through the model intact, while low content fidelity means the AI cites you while getting the details wrong. Term coined by Adam Malamis, founder of Gander.
Content Freshness
How recently content was published or meaningfully updated, which matters because answer engines favor current sources and stale pages quietly teach the models stale facts.
Context Window
The amount of text a model can consider at once, so content that states its key facts early and clearly is more likely to survive the model’s limited attention.
Crawlability for AI
How easily AI bots can access and parse your pages. Many AI crawlers do not run JavaScript, so content that only appears after scripts load may be invisible to them.
E-E-A-T
Short for Experience, Expertise, Authoritativeness, and Trustworthiness, the signals Google uses to describe credibility. The framework now doubles as a rough guide to what makes content worth citing for an AI.
Embeddings
Numerical representations of meaning that let AI systems judge how close a piece of content sits to a given query. Relevance in AI search is measured this way rather than through keyword matches.
Entity
A distinct thing that AI systems recognize and track, such as a company, product, or person. Establishing your brand as a clearly defined entity is the foundation of both visibility and fidelity.
Entity Confusion
When an AI mixes your brand up with a similarly named company, an acquired product, or an older version of your business, producing a classic and costly brand fidelity failure.
Fine-Tuning
Additional training that adapts a general model to a specific task or field, sharpening its behavior without rebuilding the entire model.
Generative AI
AI that creates new content such as text, images, code, and audio rather than simply classifying what already exists. Answer engines are generative AI applied to search.
Generative Engine Optimization (GEO)
The practice of earning presence in AI-generated answers by being mentioned, cited, and described accurately when answer engines respond to the questions your buyers ask. GEO is to AI search what SEO was to Google.
Grounding
Tying an AI answer to specific retrieved sources rather than to the model’s general training. Grounded answers arrive with citations, while ungrounded answers arrive with risk.
Hallucination
Confident, believable, and false information generated by an AI, including claims about your company that no source ever made. Monitoring for hallucinations has become part of modern brand protection.
Information Gain
The unique data, original research, or firsthand insight a page adds beyond what is already published. AI systems increasingly favor sources that say something new over sources that repeat what everyone else has said.
Knowledge Decay
An AI’s habit of repeating outdated facts absorbed during training, such as old pricing, retired products, or positioning you abandoned years ago. Fresh, crawlable content is the antidote.
Knowledge Graph
The structured map of entities and relationships that search and AI systems consult to understand who you are, what you make, and how you fit into your market.
Large Language Model (LLM)
The technology underneath every answer engine, a model trained on enormous amounts of text that predicts language and, in doing so, writes the answers your buyers read.
llms.txt
A proposed plain text file at a site’s root that points AI systems toward the pages that represent you best, working as a curated map for machine readers.
Machine Learning
The method behind modern AI, in which systems learn patterns from data rather than following rules written by hand.
Mention
Any instance of your brand being named in an AI-generated answer, with or without a link. Mentions are the raw material of AI share of voice.
Multimodal AI
Models that understand and produce more than text, including images, audio, and video. Your charts, screenshots, and videos increasingly shape the answers written about you.
Prompt
The question a user types or speaks to an AI system. In AI search measurement, the prompt is the unit of analysis, playing the role a keyword played in traditional SEO while being longer and far more conversational.
Prompt Engineering
The craft of designing inputs that draw better outputs from AI systems. In marketing, it also describes building the prompt sets used to audit how the models see your brand.
Prompt Tracking
Running a defined set of prompts across AI platforms on a regular schedule to measure visibility, citations, sentiment, and fidelity, and to catch changes whenever the models update.
Query Fan-Out
An AI’s habit of quietly expanding one user prompt into many background searches before it answers, which means your content competes in searches no human ever typed.
Retrieval-Augmented Generation (RAG)
The architecture behind AI search, in which the system retrieves current documents from the web and then generates its answer from them. RAG is why fresh, well structured content can influence answers within days rather than waiting on training cycles.
Search Gap
The distance between the queries AI engines actually run on a topic and the content you have published to answer them. Closing search gaps means writing for the questions the machine asks, not just the ones people type.
Semantic Chunking
Writing in self contained sections that each make sense on their own, because retrieval systems pull passages rather than whole pages. A section that can stand alone can be cited alone.
Semantic Search
Retrieval based on meaning rather than exact words, powered by embeddings. A page can surface for a question that shares none of its vocabulary and miss for one that shares all of it.
Sentiment
Whether AI answers portray your brand in a positive, negative, or neutral light. Because the models compress opinions gathered from across the web, sentiment in AI answers is a distilled reading of your reputation.
Share of Voice
How often your brand is mentioned in AI answers compared with all competing brands across the same set of prompts, making it the core competitive metric of AI search.
Structured Data (Schema Markup)
Machine readable labels added to a page’s code that tell AI systems exactly what they are looking at. Common types cover organizations, products, frequently asked questions, and authors.
Third-Party Signals
What trusted independent sources such as review sites, trade press, forums, and analysts say about you. AI answers lean heavily on outside confirmation, so your presence beyond your own site shapes the answer as much as your presence on it.
Token
The small unit of text that language models read and write, equal to roughly three quarters of a word. Model pricing, speed, and memory are all measured in tokens.
Topical Authority
Deep, interlinked coverage of a subject that convinces search engines and AI models you are a definitive source, making your pages more likely to be retrieved and cited.
Training Data Cutoff
The date after which a model’s built-in knowledge stops. Anything you have launched or changed since that date exists for the model only if retrieval finds it on the live web.
Transformer
The neural network design behind modern language models and the T in GPT, which lets a model weigh the relationships between all the words in a passage at once.
Unbranded Prompt
A prompt about a problem or category that names no company. Unbranded prompts are where buyers form their shortlists and where visibility is won or lost.
Visibility Rate
The share of tracked prompts in which your brand appears at least once, giving you a single number for how findable you are in AI search right now.
Zero-Click Search
A query resolved entirely on the answer page, with no click through to any website. AI search is overwhelmingly zero-click, so influence over the answer itself now matters more than the visit.