Glossary

AI visibility glossary

By Mandel Sherrod · Updated October 19, 2026

Thirty terms you'll run into when working on AI search and AI recommendations, each defined in a sentence or two, with no jargon used to explain jargon.

Plain-language definitions for the terms that come up when you work on being found and recommended by AI assistants and search engines. Where a term comes from a specific company, we say so and link to the source when we can.

AEO (answer engine optimization): Making your content the one an answer engine uses when it replies with a single answer instead of a list of links. The term predates today's AI assistants (it grew up around featured snippets and voice search) and is now used almost interchangeably with GEO. See GEO vs SEO.
AI Mode: Google's conversational search experience, where a question gets a longer AI-written answer with links and follow-up questions. Google says the same fundamentals as regular Search apply, with no special requirements to appear.
AI Overviews: The AI-written summaries Google shows at the top of some search results, with links to sources. A page has to be indexed and eligible to show a snippet in Google Search to be used. Controls like nosnippet limit what Google can show from a page.
AI visibility: How often, and how accurately, AI assistants and AI search features mention, recommend and cite a business when people ask relevant questions. It's usually measured by sampling answers to a fixed set of questions (see mention rate and share of voice).
Answer engine: Any system that responds to a question with a composed answer rather than a ranked list: ChatGPT, Perplexity, Gemini, Claude, Copilot, Google's AI Overviews and AI Mode, and voice assistants.
Brand mention: Your business being named in an AI answer or on a third-party page, with or without a link. In AI search, unlinked mentions on sites the engine reads can still shape whether and how you're described.
Canonical URL: The preferred address for a page when several URLs show the same content, declared with <link rel="canonical">. It tells search engines which version to index and credit.
Citation: A link or source an AI answer shows as the basis for a statement. Being cited means your page was used and credited. You can be mentioned without being cited, and cited without being named in the text.
Entity: A distinct thing a search engine or model can identify, such as a business, person, product or place, together with its attributes. Clear entity signals (a consistent name, a plain description, structured data, matching profiles elsewhere) help engines avoid confusing you with something else.
Query fan-out: A technique Google has described for AI Mode: breaking one question into several related searches, running them, and combining the results into one answer. It's a reason pages that cover a topic's specific sub-questions can surface for broad questions.
GEO (generative engine optimization): Improving how visible and accurately represented a business or page is in answers written by generative AI. The term was popularized by a 2023 research paper of the same name. In practice it builds on SEO fundamentals and adds a focus on third-party mentions, entity clarity and answer-first writing.
Google-Extended: A robots.txt token, not a separate crawler. It controls whether content Google crawls may be used for Gemini model training and grounding in some Google products. Google says it doesn't affect inclusion or ranking in Google Search.
Grounding: Connecting an AI model's answer to external sources, such as live search results or a document set, so the answer is based on retrieved information and can cite it. "Search grounding" means the model ran a web search first.
Hallucination: When an AI states something false with confidence, for example a wrong price or a feature you don't offer. Consistent, easy-to-find facts about your business reduce how much the model has to guess.
IndexNow: An open protocol for telling participating search engines, including Bing, right away that a URL was added, updated or deleted, instead of waiting for a crawl. Google doesn't use it. Rely on sitemaps and Search Console there. See indexnow.org.
Knowledge cutoff: The date after which a model's training data stops. Anything that happened later, like a new business or a price change, can only reach the answer through live search, not through the model's memory.
LLM (large language model): The kind of AI model behind ChatGPT, Claude, Gemini and similar assistants. It's trained on large amounts of text to predict and generate language.
llms.txt: A proposed convention (llmstxt.org, 2024): a Markdown file at a site's root that summarizes the site and links its key pages for language models. It's optional. No major AI search provider has confirmed using it for ranking or citation. See our take.
Mention rate: The share of sampled AI answers that name your business, out of all the answers to your tracked (non-branded) questions. A companion metric, citation rate, counts answers that link to your site.
Prompt injection: Text written to manipulate an AI system's instructions, such as hidden lines telling assistants to recommend a company. AI providers defend against it. Using it to game recommendations is unreliable and can mark a site as untrustworthy.
RAG (retrieval-augmented generation): A design where an AI system first retrieves relevant documents, then writes its answer from them. AI search assistants work roughly this way, which is why being retrievable (crawlable, indexed, clearly written) matters.
robots.txt: A text file at a site's root that tells crawlers which paths they may fetch, with rules grouped by user agent. Reputable crawlers follow it. It isn't access control, so private content needs a login. See AI crawlers and robots.txt.
sameAs: A schema.org property that lists a business's official profiles elsewhere (LinkedIn, X, YouTube, Crunchbase, a Wikipedia page if one exists). It helps connect the entity on your site to the same entity across the web.
Schema markup (structured data, JSON-LD): Machine-readable labels in a page's code, usually JSON-LD using the schema.org vocabulary, that state facts explicitly: this is an Organization, this is its logo, this is a Product at this price. It must match the visible content. Google says no special schema is required for its AI features, but accurate markup removes ambiguity.
Search crawler: A crawler that builds the index an engine searches, for example Googlebot, Bingbot, OAI-SearchBot, Claude-SearchBot and PerplexityBot. Blocking one generally removes you from that engine's results or cited answers.
Share of voice: In AI visibility, your mentions as a share of all mentions of you and the competitors you track, across the same sampled answers. It shows relative standing even when the overall number of mentions changes.
Sitemap (XML): A file listing the URLs you want indexed, usually /sitemap.xml, referenced from robots.txt and submitted in Google Search Console and Bing Webmaster Tools.
Training crawler: A crawler that collects public content that may be used to train AI models, for example GPTBot (OpenAI), ClaudeBot (Anthropic) and CCBot (Common Crawl). Major AI companies document these separately from their search crawlers.
User-initiated fetcher: An agent that loads a specific page because a person asked an assistant to, for example ChatGPT-User, Claude-User and Perplexity-User. OpenAI and Perplexity note that robots.txt may not apply to these fetches.
Zero-click search: A search that ends without a click to any website, because the answer was shown on the results page or in an AI summary. It's one reason being named in the answer matters, not only being linked.

Want this done for you?

We make daily, human-reviewed improvements so AI assistants and search engines can find you, trust what they read, and recommend you by name. A report lands every Monday.

Get started