AI Visibility Glossary

An AI assistant is a conversational tool like ChatGPT, Gemini, Claude, Perplexity, or Copilot that answers questions directly in natural language. Instead of returning a list of links, it gives a single synthesized reply, which makes it the new front door to information and to brand recommendations.

AI discovery is how buyers now find brands through AI answers instead of browsing links. When a buyer asks an assistant for options, the model names a short shortlist, so brands absent from that answer are missed early, often before the buyer ever visits a website.

AI search is search that returns a synthesized, written answer instead of a list of links. Surfaces like ChatGPT, Perplexity, and Google AI Overviews read across sources and reply directly, which is the exact surface that GEO and AEO work to influence.

AI visibility is how often and how prominently your brand appears in the answers AI assistants give to relevant questions. It is the AI-era version of search visibility: instead of measuring rankings on a results page, it measures your presence inside ChatGPT, Gemini, Perplexity, Claude, and Grok answers.

An answer engine is any system that returns a direct answer instead of a list of links, including AI assistants, featured snippets, AI Overviews, and voice search. Optimizing to be that answer, or the source it cites, is the discipline called answer engine optimization (AEO).

Answer Engine Optimization (AEO) is the practice of optimizing your content to be the direct answer a machine gives, whether that is an AI assistant, a featured snippet, or a voice result. Instead of earning a click, you earn the response itself, ideally with your brand named as the source.

A buyer-intent prompt is a question someone asks an AI assistant when they are close to a purchase decision, such as "best CRM for a small agency" or "alternatives to Salesforce." These are the prompts where being mentioned directly influences a shortlist, which makes them the ones worth tracking first.

Conversational search is querying in natural language across multiple turns, refining and following up instead of starting over with new keywords. Buyers narrow toward a decision by asking the assistant to compare, filter, and justify, which changes what content earns a place in the answer.

Generative AI is any AI system that creates new content, such as text, images, code, or answers, rather than only classifying or retrieving existing data. In marketing it matters because generative assistants now write direct answers to buyer questions, changing how brands are discovered and recommended.

Generative Engine Optimization (GEO) is the practice of getting your brand mentioned, cited, and recommended inside AI-generated answers from tools like ChatGPT, Gemini, Perplexity, Claude, and Grok. Where classic SEO competes for a rank on a results page, GEO competes to be part of the answer itself.

A large language model (LLM) is an AI system trained on huge amounts of text that generates language by predicting the most likely next words. LLMs power assistants like ChatGPT, Gemini, and Claude, letting them answer questions in fluent prose instead of returning a list of links.

LLM optimization (LLMO) is the practice of improving how often and how favorably your brand appears in answers from large language models like ChatGPT, Gemini, and Claude. It is largely a synonym for GEO and AEO, since the field has several overlapping names for the same craft.

A prompt is the natural-language input a person gives an AI assistant, such as "best CRM for a small agency." In AI visibility work the exact phrasing of buyer prompts is the unit of measurement, because whether your brand appears depends on the specific question asked.

Search engine optimization (SEO) is the practice of improving how well your pages rank in search engines like Google through crawlable sites, authority, and relevant content. It is the baseline GEO builds on, since the same fundamentals of access and trust also help AI engines find and cite you.

Zero-click search is when the answer appears directly on the surface, so the user never clicks through to a website. AI answers, featured snippets, and AI Overviews all satisfy the query in place, which makes being the cited and named source matter more than a raw ranking.

An AI citation is when an AI assistant links to or names your website as a source for the answer it gives. Unlike a passing brand mention, a citation credits your page as the evidence behind a claim, which drives referral traffic and signals that the engine trusts your content.

AI referral traffic is website visits that arrive from AI assistants like ChatGPT, Perplexity, Gemini, and Copilot, when a user clicks a link inside an AI answer. It appears in analytics as referrals from those domains and serves as downstream proof that your content is being cited, not just mentioned, in AI responses.

An AI visibility score is a composite metric that rolls presence, prominence, citation, and sentiment across multiple AI engines into a single number. It is a rollup meant to show your overall standing in AI answers at a glance, not a ranking position, so it is most useful as a trend line rather than an absolute grade.

A brand mention is any time an AI assistant names your brand in an answer, whether or not it links to you. Mentions are the base unit of AI visibility, but a mention alone does not mean you were recommended, so it is worth separating being named from being endorsed.

Citation share is your portion of the source links an AI engine cites across a prompt set, measured against competitors. It counts how often your pages are used as the sources behind an answer, which is distinct from being named in the text. A brand can be mentioned without being cited, and cited without being named.

Competitive benchmarking is the practice of measuring your AI visibility against a named set of rivals on the same prompts and the same engines. Instead of asking whether you appear, it asks whether you are winning or losing the category. It turns an isolated score into a relative standing you can act on.

Mention rate is the percentage of tracked prompts where your brand is named at all in the AI answer. It is the baseline presence metric, measured before any judgment of quality, position, or sentiment. If forty of one hundred tracked prompts name you, your mention rate is 40 percent. It tells you whether you show up, not how well.

Prominence measures how prominently your brand is named inside an AI answer: whether you are the first recommendation, one option in a mid-list, or an afterthought at the end. It captures the quality of a mention rather than its existence, because being named first carries far more influence over the reader than a passing note.

A prompt gap is a buyer question where your brand should appear in the AI answer but does not, while competitors do. Prompt gaps are the most actionable finding in an AI visibility audit, because each one is a specific, winnable opportunity rather than a vague score.

A prompt set is the curated list of prompts you track to measure AI visibility, sometimes called a prompt library. Every visibility metric is calculated against it, so the set defines what you are actually measuring. A representative, buyer-intent-weighted set yields honest numbers, while a careless one produces confident but meaningless results.

Recommendation rate is the percentage of tracked prompts where an AI answer actively recommends your brand, not merely names it. It sits above mention rate because it requires a positive, endorsing context. A brand can be mentioned in passing or as a runner-up without ever being the answer the assistant tells the reader to choose.

Sentiment is how favorably an AI assistant describes your brand when it mentions you: positive, neutral, or negative. Two brands can be mentioned equally often, yet one is called "the best value" and the other "limited and pricey." Sentiment captures that difference, which raw mention counts miss.

AI share of voice is the percentage of relevant AI answers that mention your brand, measured against your competitors. If ten buyer prompts return answers and you appear in four, your share of voice is 40 percent. It turns raw mentions into a competitive standing you can track over time.

Brand authority is the overall credibility a brand accumulates across everything published about it, on its own sites and on earned sources. It is the compound signal of being a trusted, well-recognized entity, and it is one of the strongest inputs into whether AI engines feel safe recommending you.

Co-citation is being mentioned alongside recognized leaders in your category across many sources. When trusted pages repeatedly name you in the same breath as established players, engines learn to associate you with that group, which pulls you into the consideration set for related questions.

E-E-A-T stands for Experience, Expertise, Authoritativeness, and Trust, the quality framework Google uses to judge whether content and its source are credible. The same signals that raise E-E-A-T also make AI engines more comfortable trusting and citing you, so it maps closely onto AI visibility.

An entity is a distinct thing an AI or search engine recognizes and reasons about: a company, product, person, or concept, along with everything it knows about it. AI engines recommend entities, not keywords, so being a clearly defined entity is the foundation of AI visibility.

Entity clarity is how consistently and confidently AI engines understand who you are, what you do, and who you serve. High entity clarity means every source agrees on your identity and category, so a model recommends you without hesitation. Low clarity means it is unsure, so it stays quiet or names a competitor.

Entity disambiguation is the step where a system resolves an ambiguous name to the one specific entity it means, choosing your brand over a same-named company, product, or ordinary word. When it resolves incorrectly, mentions of you get credited to something else and you lose recommendations you earned.

A knowledge graph is a structured network of entities and the relationships between them that search and AI systems use to reason about the world. Instead of storing loose keywords, it stores facts like which company makes which product and competes with whom, which is how engines decide what to recommend.

A knowledge panel is the box of structured facts a search engine shows about a recognized entity, drawn from its knowledge graph. It signals that the engine has a confident, canonical understanding of who you are, which is the same understanding AI assistants draw on when they decide whether to recommend you.

Named Entity Recognition (NER) is the process by which a system detects and classifies named things in text, such as companies, products, and people. It is how an engine spots that a string of words refers to your brand, which is the first step before it can associate you with a category or recommend you.

Topical authority is the depth and breadth of credible coverage a brand has across an entire subject, not just one page. When you address a whole topic cluster thoroughly and consistently, engines grow confident that you understand the field, which makes them more willing to cite and recommend you within it.

Wikidata is a free, open, structured knowledge base that stores facts about entities in machine-readable form. Search engines, knowledge graphs, and many AI models draw on it, so an accurate Wikidata entry helps systems recognize your brand as a well-defined entity and reason about it correctly.

An alternatives page is content built around 'alternatives to [competitor]', positioning your brand for buyers actively looking to switch. Done with genuine value, it wins high-intent switching prompts because AI assistants asked for alternatives to a product pull from pages that fairly lay out the options.

Answer-first content leads with a direct, self-contained answer to the reader's question, then explains and supports it below. Because AI engines lift short passages to build their answers, opening with the answer makes your content easy to quote and more likely to be cited.

Comparison content is a page that weighs two or more options against each other, such as an 'X vs Y' or 'alternatives to X' article. It feeds consideration-stage prompts because AI assistants asked to compare products draw on pages that already lay out the differences clearly and fairly.

Content freshness is how current and recently updated a page is. It acts as a trust and retrieval signal because AI answers often favor recent, maintained content, especially for fast-moving topics. Keeping pages honestly up to date makes them more likely to be retrieved and cited than stale ones.

Digital PR is the active practice of earning mentions, coverage, and links on trusted publications and sites. It directly grows AI visibility because the third-party citations and entity signals it creates are the evidence engines use to decide which brands are credible enough to name and recommend.

Earned media is third-party coverage you did not pay for or own, such as press articles, independent reviews, and organic mentions. It outweighs owned content for AI visibility because an engine trusts what others say about you more than what you say about yourself, which is exactly the signal it needs to cite you.

FAQ content is a set of question-and-answer blocks, often marked up with FAQ schema, that answers the real questions people ask about a topic. Its format is uniquely liftable by answer engines because each self-contained question and short answer maps directly onto the way AI assistants generate responses.

Listicle content is an article structured as a ranked or grouped list, most often a 'best X for Y' roundup. It dominates buyer-intent AI answers because assistants asked to recommend options can lift a well-scoped, clearly reasoned list almost directly, so honest listicles are one of the most citable formats in GEO.

A long-tail query is a specific, lower-volume search or prompt, usually longer and more detailed than a broad head term. AI assistants excel at these because they synthesize an answer to the exact question, and a focused brand can win the long tail where it could never outrank giants on generic terms.

Review signals are the ratings and reviews about a brand on platforms like G2, Capterra, and Trustpilot. They shape which brands AI assistants recommend in a category because these sites are trusted, structured, and heavily cited sources that an engine reads as independent evidence of quality and fit.

A semantic core is the defined set of topics, terms, and entities that establish what your brand is relevant for. It guides which content you create and how you connect it, so that search and AI systems build a clear, consistent understanding of your area of authority.

AI crawlers are the bots AI companies use to read the web, either to train models or to fetch live pages when answering a question. Examples include GPTBot, OAI-SearchBot, PerplexityBot, and ClaudeBot. If you block them, your content cannot be read, cited, or recommended by those systems.

Chunking is the process of splitting content into smaller passages so it can be embedded, stored, and retrieved individually. AI retrieval works at the passage level, so clear, self-contained sections are far more likely to be pulled and quoted than a single undivided wall of text.

A context window is the maximum amount of text, measured in tokens, that a model can consider at once, including the prompt, any retrieved sources, and its own answer. It is finite, so it limits how many sources can ground a single answer and how much of each one the model actually reads.

GPTBot is OpenAI's web crawler, identified by the user-agent name GPTBot. It fetches public pages that OpenAI may use to improve its models. You control it in robots.txt by allowing or disallowing that user agent, and the choice affects how well your content is understood by OpenAI systems.

Grounding is when an AI model bases its answer on retrieved, verifiable sources rather than its own memorized parameters. A grounded answer can point to where each claim came from. Grounding is what makes AI answers citable, and it is the mechanism that lets your content become part of the response.

Indexation is whether your pages are present in the index that an answer engine retrieves from. Many AI assistants draw on a search index, such as Bing behind ChatGPT, so if your page is not indexed there, it cannot be retrieved or cited no matter how good it is.

llms.txt is a proposed plain-text file at the root of your site that gives AI models a clean, curated map of your most important content in Markdown. Like robots.txt for crawlers or sitemap.xml for search, it is a convention meant to help LLMs find and understand your key pages.

The Model Context Protocol (MCP) is an open standard for connecting AI assistants to live data and tools through a consistent interface. Instead of a model relying only on its training, an MCP server lets it query real systems in real time. Rankry offers an MCP server that exposes report data to AI clients.

Passage ranking is when a system scores and ranks individual passages of a page rather than the page as a whole. One strong, directly relevant section can be retrieved and quoted even if the overall page is not the most authoritative, which shifts optimization toward the section level.

Query fan-out is when an engine takes one user question and expands it into many related sub-queries, runs them, and merges the results into a single answer. Because your content can be pulled in through any of those sub-queries, broad coverage of a topic matters more than matching one exact phrase.

Retrieval-augmented generation (RAG) is the technique where an AI model, before answering, retrieves relevant documents from a search index or the live web and uses them as source material. It is why AI answers can cite current pages, and why being retrievable and citable matters for your visibility.

robots.txt is a plain text file at the root of a domain that tells crawlers, including AI crawlers, which paths they may or may not fetch. It works by user-agent name and path rules. A single wrong Disallow line can silently keep your pages out of the index an answer engine draws from.

Semantic search retrieves content by meaning rather than by matching exact keywords. It compares the intent behind a query with the meaning of stored passages using vector embeddings, so a page can be found even when it never uses the searcher's exact words. It underpins how AI systems locate the material they cite.

Server-side rendering means the server sends a fully built HTML page, with the content already in it, rather than an empty shell that JavaScript fills in later. Because many AI crawlers do not run JavaScript, SSR keeps your content readable to them, avoiding a common silent cause of invisibility.

Structured data is machine-readable code, usually schema.org markup in JSON-LD, that states explicitly what a page and its entity are: an organization, a product, an FAQ, an article. It removes ambiguity for engines, making your content easier to parse, trust, and quote in AI answers.

A vector database stores text as numeric vectors, or embeddings, and finds the ones closest in meaning to a query. It is the retrieval engine inside many AI systems, holding chunked content so that a relevant passage can be pulled fast when an answer is being assembled.

A vector embedding is a list of numbers that represents the meaning of a piece of text, produced by a model so that similar meanings sit close together in a shared mathematical space. Embeddings are what let AI systems find content by meaning rather than exact words, powering semantic search and retrieval.

AI Overviews are Google's AI-generated answers that appear at the top of search results, summarizing information and citing a few sources above the traditional blue links. They push classic results down the page, so being one of the cited sources is now a distinct visibility goal alongside ranking.

Bing is Microsoft's web search engine and, just as importantly, the search index behind several AI products. ChatGPT's web search and Microsoft Copilot both draw on Bing to retrieve and cite live pages. That makes being indexed and crawlable in Bing a practical, often overlooked, lever for AI visibility.

ChatGPT is OpenAI's AI assistant and the most widely used AI chat product. People ask it questions and get one synthesized answer instead of a list of links, which makes it a default front door to product recommendations. It answers from training data by default and can also search the live web when a query needs current information.

ChatGPT Search is ChatGPT's ability to fetch live web pages and answer with current information, citing the sources it used. It relies on a search index and answer-time crawlers rather than only training data, which is why crawlability and being indexed decide whether your brand can appear.

Claude is Anthropic's AI assistant, known for careful, long-form reasoning and a large context window. It is used both as a consumer chat product and widely through its API inside other tools. For brands, Claude is one of the five major engines to monitor, since it can recommend a different set of options than ChatGPT or Gemini for the same question.

Google AI Mode is a conversational search experience in Google that lets users ask a question, read an AI-generated answer, and keep asking follow-ups in the same thread. It goes further than a single AI Overview by supporting a back-and-forth dialogue, often breaking a complex query into many sub-searches behind the scenes.

Google Gemini is Google's family of AI models and its consumer AI assistant. Gemini powers a standalone chat app and is woven into Google products, and Google's models also underpin AI-generated answers in Search such as AI Overviews. For brands, it is one of the major engines to track alongside ChatGPT, Claude, Perplexity, and Grok.

Grok is xAI's AI assistant, integrated with the social platform X. Alongside general chat and reasoning, its distinguishing angle is access to real-time posts and discussion on X, which lets it draw on very recent social conversation. For brands, it is one of the major engines to track because it can answer differently than models focused on the open web.

Microsoft Copilot is Microsoft's AI assistant, spread across Bing, Windows, Edge, and Microsoft 365. It answers questions conversationally and draws on the Bing search index for live web results with citations. Because it is embedded in tools that many workplaces already use, Copilot gives Microsoft broad, especially enterprise, reach for AI answers.

Perplexity is an AI answer engine that responds to questions with a short synthesized answer and prominent numbered citations to the sources it used. Because it foregrounds sources by design, it is one of the clearest surfaces for seeing which pages an AI actually pulled from, and a high-value place to earn visible citations.

AI brand safety is the practice of managing how AI assistants portray your brand. It covers misinformation, outdated claims, and unfavorable framing that can reach buyers as fact. Because you cannot edit the model, the work is monitoring what assistants say and correcting the sources they read so the picture stays accurate.

Attribution is whether and how an AI answer credits the sources behind it. When an assistant links or names the page a claim came from, that citation is attribution in action. It matters twice over: it builds trust in the answer, and it is the path by which AI sends referral traffic back to your site.

A hallucination is when an AI model states something false with full confidence, presenting invented facts as if they were true. For a brand it is a real risk: a model can misdescribe your pricing, features, or founding story. Grounding answers in retrieved, trustworthy sources is the main way to reduce it.

A knowledge cutoff is the date a model's training data ends. The model has no built-in awareness of anything that happened after it. Newer facts can only reach the model through live retrieval at answer time, which makes being crawlable and current urgent for any brand that changes.

Model bias is a systematic skew in what an AI model tends to say or recommend. In brand terms it often shows up as incumbent bias, a lean toward well-known names the model has seen most. Challengers counter it by building the clear, credible, repeated signals that make a model confident enough to name them.

Training data is the large body of text a model learned from during training. It shapes what the model knows by default, but it is frozen at a point in time and never complete. That is why models supplement it with live retrieval at answer time, pulling fresh sources the training data never contained.