AI search comes with a lot of new words. This glossary explains them in plain language, one term at a time.
Each term page cites its official sources, such as Google, OpenAI or Anthropic. Some terms already have a full page on GEO-Score; for those, we link straight to it.
What happens inside an AI tool between your question and its answer.
An AI hallucination is an answer that sounds confident and believable but is wrong or not supported by any source.
Chunking means cutting long documents into smaller pieces of text before they are indexed, so a retrieval system can return just the passage that answers a question.
The context window is the amount of text a language model can look at and use while it writes an answer, a kind of working memory.
An embedding is a long list of numbers that represents the meaning of a piece of text, so a computer can measure how related two texts are.
Grounding is when an AI model looks up information in real sources, such as a search index, while it writes an answer, instead of relying only on what it learned in training.
Hybrid search combines keyword matching, often with a formula called BM25, and meaning-based vector search, and then merges the results.
A knowledge cutoff is the date after which a model’s training data stops, so the model knows nothing newer unless it looks it up.
An AI model trained on huge amounts of text that can read and write language.
Explained in: What is a large language model?
Reranking is a second sorting step in a retrieval system: it first grabs many possible passages quickly, then a more careful model re-orders them and keeps only the best few.
Retrieval-augmented generation (RAG) is a way of building AI answers in two steps: first find relevant text (retrieval), then let a language model write the answer using that text (generation).
Semantic search tries to understand what a question means, so it can find useful pages even when they don’t use the exact same words.
Tokens are the small pieces of text a language model reads and writes: whole words, parts of words, single characters or even bytes.
Training data is what a model learned before release. Retrieval is content it looks up while it answers.
Explained in: When AI models use your content
Vector search finds results by comparing embeddings, so it returns text with a similar meaning instead of only text with the same words.
The boxes, answers and reports you meet in AI search, and how to measure them.
An AI brand mention is when an AI-generated answer names your company, product or service, whether or not it also links to your website.
A link to your page shown as a source in an AI-generated answer.
Explained in: Guide: getting cited by AI
A metric used by AI visibility tracking tools: how often answers link to your site for the questions you track.
Explained in: Guide: getting cited by AI
Google's AI search mode for questions that need more exploring, reasoning or comparing, with links to sources.
Explained in: Guide: Google AI Overviews
Google's AI-written summary on the results page, with links to supporting websites.
Explained in: Guide: Google AI Overviews
AI performance reports are reports in Bing Webmaster Tools and Google Search Console that show how often your pages appear in AI-generated answers.
Visits to your site that come from links in AI tools such as ChatGPT or Perplexity.
Explained in: Guide: AI referral traffic
How often your brand shows up in AI answers compared with competitors, for the same questions.
Explained in: Guide: AI share of voice
An answer engine is a search tool that gives you a written answer to your question instead of only a list of links.
A featured snippet is a special Google search result that shows a short quote from one page first, above the link to that page.
Search where an AI writes a new answer from sources it finds, instead of only listing links.
Explained in: How AI search works
A knowledge panel is an information box Google shows when you search for a person, place, organization or thing that is in its Knowledge Graph.
Search that mixes text with images, video or voice, in the question or in the answer.
Explained in: Guide: multimodal GEO
Passage ranking is an AI system Google uses to identify individual sections, or passages, of a web page, to better understand how relevant the page is to a search.
An AI turns one question into several related searches and combines the results. Google names this technique for AI Overviews and AI Mode.
Explained in: Guide: query fan-out
Rich results are Google search results with extra features, such as images, ratings or event details, that Google can show based on structured data on the page.
Asking a question out loud to an assistant and getting a spoken answer.
Explained in: Voice search and AI
A search where people get the answer on the results page and never click through to a website.
Explained in: Guide: zero-click AI search
How machines learn who you are and what your page is about.
A clearly defined thing, like a person, company or place, that search engines and AI can recognize and connect to facts.
Explained in: Guide: entity SEO
FAQPage is a schema.org type for a web page that presents one or more frequently asked questions with their answers.
The format most sites use to add structured data, inside a script tag. Google recommends it but says all three supported formats are fine.
Explained in: Schema implementation guide
Organization schema is structured data that describes a company or other organization, such as its name, logo, website and profiles, using the schema.org Organization type.
sameAs is a schema.org property that lists web pages which clearly show the identity of the thing described, such as its Wikipedia page, its Wikidata entry or its official profiles.
Structured data that uses the schema.org vocabulary to describe a page.
Explained in: Schema validation guide
Labels in a page's code that tell machines what things are, such as a product, a price or an author.
Explained in: Guide: structured data for AI search
Wikidata is a free, collaborative, multilingual knowledge base that stores structured facts about people, places, organizations and ideas.
The bots that visit your site, who runs them and what blocking them does.
A bot that collects web pages for an AI company: for training, for search, or to answer one user.
Explained in: Guide: AI crawler access
Applebot-Extended is a robots.txt token that lets you opt out of Apple using your website content to train its generative foundation models.
Bingbot is the web crawler Microsoft uses for Bing search.
CCBot is the web crawler of Common Crawl, a non-profit foundation that builds an open repository of web crawl data anyone can access and analyze.
ChatGPT-User is the user agent ChatGPT and Custom GPTs use when they visit a web page because a person asked a question.
Claude-SearchBot is Anthropic’s crawler that navigates the web to improve the quality of the search results Claude gives its users.
ClaudeBot is Anthropic’s web crawler that collects content which could contribute to training its generative AI models.
Google-Extended is a robots.txt token that controls whether content Google crawls may be used to train Gemini models and to ground answers in Gemini Apps and Vertex AI.
Googlebot is the generic name for Google Search’s two web crawlers, Googlebot Smartphone and Googlebot Desktop.
GPTBot is OpenAI’s web crawler that collects content which may be used to train OpenAI’s generative AI foundation models.
A proposed Markdown file that gives AI tools a short overview of your site. Google says Google Search ignores it.
Explained in: Guide: llms.txt
Meta-ExternalAgent is Meta’s web crawler for uses such as training foundation AI models or improving products by indexing content directly.
OAI-SearchBot is the OpenAI crawler that surfaces websites in the search features of ChatGPT.
PerplexityBot is Perplexity’s crawler that surfaces and links websites in Perplexity’s search results.
A text file at the root of your site that tells crawlers which parts they may visit. It is a request, not a lock.
Explained in: robots.txt guide
Tags, headers and protocols that decide what search engines crawl and show.
A canonical URL is the preferred address of a page when the same or very similar content is available at several URLs.
Google's measurements of a page's loading speed, responsiveness and visual stability.
Explained in: Page speed guide
Crawl budget is the time and resources Google can devote to crawling a single site.
data-nosnippet is an HTML attribute that tells Google not to use a specific part of a page in its search snippets.
hreflang is an annotation that tells Google which language or regional versions of a page belong together, so people get the right one.
IndexNow is a protocol for instantly telling participating search engines that a URL on your site was added, updated or deleted.
JavaScript rendering is the step where a browser runs a page’s JavaScript to build its content; text that only appears after this step is invisible to crawlers that don’t run JavaScript.
max-snippet:[number] is a robots rule that sets the maximum number of characters Google may use as a text snippet for a page.
noindex is a rule, set with a meta tag or an HTTP header, that tells search engines like Google not to show a page in their results.
nosnippet is a robots rule that tells Google not to show a text snippet or video preview for a page in its search results.
Web Bot Auth is an authentication method in which a bot signs its HTTP requests cryptographically, so a website can verify that a request really comes from that bot.
X-Robots-Tag is an HTTP response header that can carry any rule the robots meta tag can, such as noindex or nosnippet.
A file that lists the pages on your site so search engines can find them.
Explained in: Sitemap and discoverability guide
AI systems that use websites for people, and the new standards they rely on.
The accessibility tree is the browser’s simplified version of a page that keeps what matters for interaction: the roles, names and states of elements like buttons, links and form fields.
Agentic browsing is when an AI agent uses a website the way a person would: reading pages, clicking buttons and filling in forms.
An AI agent is an autonomous AI system that performs tasks on behalf of people, such as booking a reservation or comparing product specifications.
An open-source standard for connecting AI applications to outside systems, such as data sources and tools.
Explained in: Guide: MCP and AI agents
NLWeb is an open project, introduced by Microsoft in May 2025, that makes it simpler to add natural-language interfaces to websites.
The Universal Commerce Protocol (UCP) is an open-source standard that gives platforms, AI agents and businesses a common language for agentic commerce, from discovery to checkout.
WebMCP is a proposed browser API that lets web applications offer JavaScript-based tools to AI agents.
The many abbreviations people use for optimizing for AI search. Each one has its own page.
Making your brand findable and usable for AI agents that research and compare for people.
Explained in: AAO: Assistive Agent Optimization
Structuring content so answer engines can lift it out as a direct answer to a question.
Explained in: Answer engine optimization (AEO)
Classic SEO combined with optimizing for AI answers: ranking in results and getting cited.
Explained in: AI SEO
How well your brand is found, cited and represented by AI assistants and AI search.
Explained in: AI visibility
Optimizing content for AI-powered search engines that write answers instead of listing links.
Explained in: AIEO: AI Engine Optimization
Making sure AI systems understand your content correctly and can cite it.
Explained in: AIO: Artificial Intelligence Optimization
Making sure your brand is shown accurately and prominently inside AI-generated answers.
Explained in: AIRO: AI Results Optimization
Making content easy for AI search engines to find, extract and cite.
Explained in: AISEO: AI Search Engine Optimization
Helping AI-driven search engines interpret and recommend your content.
Explained in: AISO: Artificial Intelligence Search Optimization
Hands-on tactics to make a brand visible and quotable across AI platforms.
Explained in: ALLMO: Applied Large Language Model Optimization
Aiming to be the answer: in featured snippets, AI Overviews and direct AI answers.
Explained in: ASO: Answer Search Optimization
Improving how often and how well your brand appears in answers from AI models.
Explained in: GAIO: Generative AI Optimization
Optimizing content to appear in AI-generated search results built from several web sources.
Explained in: GAISO: Generative AI Search Optimization
Optimizing content so it appears in answers written by AI, such as ChatGPT, Perplexity and Google AI Overviews.
Explained in: GEO: Generative Engine Optimization
Making content ready to be cited by AI search engines, so your brand becomes part of the answer.
Explained in: GSO: Generative Search Optimization
An informal name for optimizing content for large language models, close in meaning to GEO.
Explained in: Guide: LLM SEO
Writing and structuring content so large language models can understand, extract and cite it.
Explained in: LLMO: Large Language Model Optimization
The basics of making content understandable and correctly represented by AI systems.
Explained in: OAI: Optimization for AI
One approach for both classic search engines and AI-powered search.
Explained in: SAIO: Search AI Optimization
Being findable wherever people search: Google, AI assistants, social media and marketplaces.
Explained in: SEvO: Search Everywhere Optimization
Optimizing the whole search journey, from typing a question to finding a good answer.
Explained in: SXO: Search Experience Optimization
The signals a GEO-Score report measures. Each one has its own guide.
Whether AI crawlers may reach a page, based on robots.txt and server responses.
Explained in: AI bot access guide
A combined score for technical AI-readiness signals, such as schema, structure and freshness.
Explained in: AI optimization guide
Whether a section fully answers a question on its own.
Explained in: Answer completeness guide
How easily a passage can be quoted on its own in an AI answer.
Explained in: Citability guide
Whether a page cites outside sources for its claims.
Explained in: Citations and sources guide
How well a page covers a topic's subtopics and follow-up questions.
Explained in: Comprehensiveness guide
Words that connect ideas between sentences and paragraphs.
Explained in: Content cohesion guide
How recently content was published or meaningfully updated, and whether that date is shown.
Explained in: Content freshness guide
How clearly headings, lists and sections organize a page.
Explained in: Content structure guide
Whether the format of a page, such as a list, table or how-to, fits the question it answers.
Explained in: Content type matching guide
Experience, Expertise, Authoritativeness and Trust: signals that show who is behind content and why to trust it.
Explained in: E-E-A-T guide
How many checkable facts, numbers and named sources a text contains.
Explained in: Factual density guide
Whether images have text descriptions that machines and screen readers can read.
Explained in: Image alt text guide
A database of entities and how they relate, like the one Google uses to understand people, places and things.
Explained in: Knowledge graph guide
How fast a page responds and loads.
Explained in: Page speed guide
How well content fits the way different AI platforms find and show sources.
Explained in: Platform specificity guide
How easy a text is to read, measured with a readability score.
Explained in: Readability guide
Whether a page's structured data is present and valid.
Explained in: Schema validation guide
How clearly a page states what it is about, with named things and plain relationships.
Explained in: Semantic clarity guide
Related words and concepts that show depth on a topic.
Explained in: Semantic keywords guide
Whether crawlers can find a page through your sitemap and internal links.
Explained in: Sitemap and discoverability guide
How completely a whole site covers a subject, not just one page.
Explained in: Topical authority guide
GEO-Score checks one page on 22 signals, from AI crawler access to structured data, and shows what to fix first.
Analyze a page