This AI search glossary defines the terms you will meet when you work on visibility in AI answers, from AEO, GEO and LLMO to crawlers, citations, retrieval and structured data. Each definition is short and written in plain English, and where we have a full guide or a free tool on the topic, the entry links to it. Terms are listed alphabetically, so use the letters below to jump straight to the one you need.
Jump to: A · B · C · D · E · F · G · H · I · J · K · L · M · N · O · P · Q · R · S · T · U · V · Z
A
AEO (answer engine optimization)
Answer engine optimization is the practice of making a brand and its content easy for AI assistants and answer features to find, understand and quote, so that the brand is mentioned and cited in their answers. The term was used for featured snippets and voice assistants before generative AI arrived, and it now mostly refers to answers from tools such as ChatGPT, Perplexity, Gemini and Google AI Overviews. It overlaps almost completely with GEO. Read the full guide: What is answer engine optimization (AEO)?
AI agent
An AI agent is an AI system that takes actions for a user instead of only answering, for example browsing several sites to compare products, filling in a form or starting a booking. Agents read pages much as crawlers and fetchers do, so accessible pages, clear product information and working forms matter for them too. As agents become more common, more of a site’s audience may be software acting on a person’s behalf.
AI crawler
An AI crawler is an automated bot run by an AI company that requests pages from websites. Crawlers differ by job: training crawlers collect content for model training, search crawlers build the index that AI answers draw on, and user-triggered fetchers retrieve a page when someone asks a question that needs it. Read AI crawlers: which bots to allow, and which to block, or identify a bot with the free AI bot user-agent checker.
AI Mode
AI Mode is a conversational search experience in Google Search in which the answer is written by AI and the user can keep asking follow-up questions. Google has described it as using a query fan-out technique, running many related searches behind a single question and combining what it finds. Because it links to web pages as sources, being crawlable and quotable matters for appearing in it.
AI Overviews
AI Overviews are the AI-generated summaries Google shows at the top of some search results, with links to the pages they draw on. They grew out of Google’s earlier Search Generative Experience experiment. For many informational searches an AI Overview is the first thing a person reads, which is why being cited in one has become a goal in its own right.
AI referral traffic
AI referral traffic is the visits your site receives when people click a link in an AI assistant’s answer, for example in ChatGPT, Perplexity, Gemini, Claude or Copilot. Many analytics tools lump these visits into direct or generic referral traffic unless you classify them by referrer, and some assistants pass no referrer at all. Check a referrer with the free AI referrer checker, or see how aeoora tracks AI referral traffic.
AI SEO
AI SEO has two common meanings. In this glossary it means optimizing for AI-driven search, such as AI Overviews, AI Mode and ChatGPT search, which makes it another name for AEO and GEO. It is also used for using AI tools to do traditional SEO work, such as drafting content or clustering keywords, so check which sense someone means. See AEO vs GEO vs LLMO vs AI SEO.
AI visibility score
An AI visibility score is a single number that summarizes how visible a brand is in AI answers across a set of tracked prompts. Vendors calculate it differently. The aeoora AI Visibility Score runs from 0 to 100 and combines mention rate, citation rate and how early the brand is named. Scores are most useful as a trend over time and against competitors measured the same way.
AIO
AIO is an ambiguous abbreviation. It can mean AI optimization, a general label for optimizing content for AI systems, or it can be shorthand for Google’s AI Overviews. When you see it in a report, proposal or contract, confirm which meaning is intended.
Answer engine
An answer engine is a system that responds to a question with a direct answer instead of a list of links. Today the term usually covers AI assistants and AI search products such as ChatGPT, Perplexity, Gemini, Claude, Copilot and Google’s AI features, while featured snippets and voice assistants were earlier examples. Test how quotable a page is for answer engines with the free answer readiness checker.
B
Bingbot
Bingbot is Microsoft’s search crawler, which builds the Bing search index. That index matters beyond Bing itself, because Microsoft Copilot grounds many of its web answers in Bing results. Blocking Bingbot, or neglecting your site’s health in Bing, can therefore reduce your AI visibility as well as your Bing rankings.
Branded prompt
A branded prompt is a question that names your brand, such as asking an assistant what a specific company does or how it compares with a rival. An unbranded prompt describes a need without naming anyone, such as asking for the best tool for a job. Unbranded prompts show whether AI engines recommend you on their own, while branded prompts show how accurately they describe you, so most tracking sets lean on unbranded prompts and use branded ones for accuracy checks.
C
CCBot
CCBot is the crawler of Common Crawl, a nonprofit that publishes a large, freely available archive of web pages. That archive has been widely used as training data for large language models, so allowing or blocking CCBot is effectively a decision about training use by many parties rather than by one company. It does not power live AI answers on its own.
Citation
A citation is a link to a source that an AI engine shows with its answer. Being cited means your page was used as evidence, and it gives the reader a way to click through, which a plain mention does not. Engines differ in how many sources they cite and how prominently, so citations are best tracked per engine. Learn more in How AI search engines choose which sources to cite and in the citations docs.
ClaudeBot
ClaudeBot is the web crawler Anthropic uses to collect public content that may be used to train its Claude models. Anthropic documents separate agents for other jobs, including fetching pages when a Claude user asks for them and indexing pages for search, so you can set different rules for each. Like other major crawlers, it is controlled through user-agent entries in robots.txt.
Co-mention
A co-mention is when your brand and another brand are named in the same AI answer. Looking at who is co-mentioned with you most often shows which brands AI engines treat as your closest alternatives, which may differ from the competitors you would pick yourself. Co-mentions are a useful input when you decide which competitors to track.
Conversational search
Conversational search is searching through a back-and-forth dialogue instead of a single typed query. The user asks a question, reads an answer and asks follow-ups that keep the earlier context, as in ChatGPT, Gemini, Perplexity or Google AI Mode. Questions in conversational search tend to be longer and more specific, which is why prompts written like real questions work better for tracking than short keywords.
D
Domain authority
Domain authority is a general name for third-party scores that estimate the strength of a website’s backlink profile, such as Moz’s Domain Authority and Ahrefs’ Domain Rating. Search engines do not use these scores directly, but they are a quick proxy for how established a site is compared with others. Compare sites with the free domain authority checker.
E
E-E-A-T
E-E-A-T stands for experience, expertise, authoritativeness and trustworthiness, a framework from Google’s Search Quality Rater Guidelines for judging the quality of content. It is not a single ranking signal or score, but it describes the qualities search systems try to reward. The same qualities, such as named authors, first-hand experience and accurate sourcing, also make content more credible to AI engines.
Embeddings
Embeddings are lists of numbers that represent the meaning of a piece of text, so that passages with similar meaning sit close together mathematically. AI search systems use them to find passages that match a question even when the wording differs, a process often called vector search. This is one reason clear, focused passages tend to be retrieved more reliably than long, mixed-topic ones.
Entity
An entity is a distinct, identifiable thing, such as a company, person, product, place or concept, as opposed to a string of words. Search engines and AI systems organize knowledge around entities and the relationships between them. Using one consistent name, description and set of facts for your brand everywhere helps machines recognize you as one entity rather than several.
F
FAQPage schema
FAQPage schema is a type of structured data that marks up a list of questions and their answers on a page. Google has narrowed which sites get FAQ rich results in its search listings, but the markup still makes question-and-answer content explicit to machines. Only mark up questions that are visibly answered on the page, and build valid markup with the free FAQ schema generator.
Featured snippet
A featured snippet is a highlighted excerpt Google shows near the top of some results pages, quoting a page that answers the query directly. It was often called position zero, and winning it was the original focus of answer engine optimization. The habits that earn featured snippets, such as a direct answer under a clear, question-style heading, still help with AI answers.
G
Generative engine
A generative engine is a search or answer system that writes a new response with a generative AI model rather than only ranking existing pages. ChatGPT search, Perplexity, Gemini, Copilot and Google’s AI Overviews and AI Mode are common examples. The term gave its name to generative engine optimization.
GEO (generative engine optimization)
Generative engine optimization is the practice of improving how often and how well your brand appears in answers produced by generative AI engines, including mentions, citations and the way you are described. The term was popularized by academic research that tested how content changes affect visibility in generative answers. In practice it overlaps almost entirely with AEO. Read What is generative engine optimization (GEO)?
Google-Extended
Google-Extended is a robots.txt control token, not a separate crawler, that lets site owners say whether content Google crawls may be used to train and ground its Gemini models outside of Search. Googlebot still does the crawling. According to Google’s documentation, the token does not affect a site’s inclusion or ranking in Google Search, and AI features in Search follow the normal Search controls.
GPTBot
GPTBot is OpenAI’s web crawler for collecting public content that may be used to train its models. OpenAI runs separate agents for other jobs: OAI-SearchBot for ChatGPT’s search features and ChatGPT-User for pages fetched when a user asks. According to OpenAI’s documentation, each is controlled separately in robots.txt, so blocking GPTBot is a training decision rather than a search one.
Grounding
Grounding is connecting an AI model’s answer to specific, retrievable sources, such as web pages or documents, instead of relying only on what the model memorized during training. Grounded answers are usually more current and easier to check, and they are where citations come from. For brands, grounding is the main route by which fresh content can shape AI answers quickly.
H
Hallucination
A hallucination is a statement an AI model presents as fact that is false or unsupported, such as a wrong price, a feature you do not offer or an invented quote. Hallucinations about brands often come from outdated or conflicting information on the web, or from gaps the model fills with guesses. The fix is usually to publish clear, consistent facts where engines look and to monitor answers so you catch errors early. See What is LLMO? for how to correct them.
I
Index
An index is the database of pages a search system has crawled, processed and stored so it can retrieve them quickly. AI search products rely on an index, whether their own, a traditional search engine’s or both, and a page that is not indexed cannot be retrieved or cited. Crawler access, a current sitemap and fast pages with their content in the HTML all help pages get indexed.
J
JSON-LD
JSON-LD, short for JavaScript Object Notation for Linked Data, is the format Google recommends for adding structured data to a page. It sits in a script tag, separate from the visible content, and describes the page with schema.org types such as Article, Organization, Product or FAQPage. Generate Article markup with the free article schema generator.
K
Knowledge cutoff
A knowledge cutoff is the point in time at which a model’s training data ends. Anything that happened or changed after it is unknown to the model unless the model retrieves it while answering. This is why an assistant may describe old pricing or a former product name, and why live, crawlable pages matter.
Knowledge graph
A knowledge graph is a structured database of entities and the relationships between them, such as a company, its founders, its products and its location. Google’s Knowledge Graph powers the knowledge panels in its search results, and open sources such as Wikidata play a similar role across the web. Consistent facts on your site, in your structured data and on trusted profiles help knowledge graphs represent your brand correctly.
L
Large language model (LLM)
A large language model is an AI model trained on very large amounts of text to understand and generate language. Models of this kind power ChatGPT, Gemini, Claude, Copilot and the AI features in search engines. What an LLM says about a brand depends on its training data and on any content it retrieves when answering.
LLMO (large language model optimization)
LLMO stands for large language model optimization: improving how large language models understand, describe and recommend your brand. It pays attention to both a model’s trained knowledge and the content it retrieves, and it is largely another name for AEO and GEO. Read What is LLMO? Large language model optimization explained.
llms-full.txt
llms-full.txt is a companion to llms.txt that contains the full text of a site’s key pages or documentation in a single Markdown file, rather than a list of links. It is popular with documentation sites that want AI tools and coding assistants to load everything in one request. Convert existing pages with the free HTML to Markdown converter.
llms.txt
llms.txt is a proposed convention for a Markdown file at the root of a website that summarizes the site and links to its most useful pages for large language models. It is a hint rather than a standard every AI company has committed to, and support varies by operator. Create one with the free llms.txt generator.
Log file analysis
Log file analysis is reviewing your web server’s access logs to see which bots and visitors requested which pages. It is the reliable way to see AI crawler activity, because bots do not run the JavaScript that browser analytics tools depend on. Combined with bot verification, it shows which AI systems read your site, how often and with what errors. See AI crawler analytics.
M
Markdown version
A Markdown version is a plain-text copy of a web page, often served at a matching address ending in .md, that leaves out navigation, scripts and styling. The idea is to give AI tools a clean version of the content that is quick to read and uses few tokens. It complements a well-structured HTML page rather than replacing it.
Mention
A mention is an AI answer that names your brand, whether or not it links to you. Mention rate, the share of tracked answers that name you, is the most basic measure of AI visibility. Mentions matter even without a click, because the reader may remember you or search for you later. See mentions and sentiment.
N
Non-determinism
Non-determinism means the same prompt can produce a different answer each time it is run. AI engines vary their wording, the brands they name and the sources they cite from one run to the next and from day to day. That is why AI visibility is measured by sampling many prompts repeatedly and reading the patterns, not by trusting a single screenshot.
O
OAI-SearchBot
OAI-SearchBot is the crawler OpenAI uses to find and index pages for the search features in ChatGPT. Unlike GPTBot, its purpose is to surface and link to websites in answers rather than to collect training data. If you want to be cited in ChatGPT search, allow it in robots.txt and make sure your firewall does not block it.
Organization schema
Organization schema is structured data that describes a company: its name, logo, website, contact details and official profiles, which are often linked with the sameAs property. It helps search engines and AI systems connect your site to the right entity and keep the facts about you consistent. Build it with the free organization schema generator.
P
Parametric knowledge
Parametric knowledge is what a model knows from its training, stored in its parameters rather than looked up at answer time. It shapes answers when the model does not search, and it can be out of date or thin for smaller and newer brands. Changing it is slow and indirect, because it only updates when a model is retrained.
People also ask
People also ask is a Google results feature that shows a box of related questions, each expanding to a short answer with a source link. The questions are a useful map of what searchers want to know around a topic. They are also good raw material for FAQ sections and for the prompts you track in AI engines.
PerplexityBot
PerplexityBot is the crawler Perplexity uses to index web pages so they can be surfaced and linked in its answers. Perplexity says it is not used to crawl content for training foundation models, and the company runs a separate agent for pages fetched at a user’s request. Blocking PerplexityBot can keep your pages out of Perplexity’s citations.
Position
Position is where your brand appears within an AI answer that names several brands, such as first, second or fifth. Being named first is generally more valuable, because readers often act on the first option they see. Position is only meaningful across many answers, since the order varies from run to run.
Prompt
A prompt is the question or instruction a user types or speaks to an AI assistant. In AI visibility tracking, prompts are the real buyer questions you monitor, such as asking for the best tool for a job or comparing two options. Generate a starting list with the free AI prompt ideas generator.
Prompt tracking
Prompt tracking is running a fixed set of prompts through AI engines on a schedule and recording whether your brand is mentioned or cited, what position it holds, which competitors appear and which sources are used. It is the AI search counterpart to rank tracking. See How to track your brand’s visibility in ChatGPT and other AI engines and the prompts and topics docs.
Q
Query fan-out
Query fan-out is when an AI search system breaks one question into several related searches, runs them, and combines the results into a single answer. A question about the best CRM for an agency might fan out into searches about pricing, integrations and reviews. Your pages compete for those sub-questions, so covering them clearly raises your chances of being used. See How AI search engines choose which sources to cite.
R
RAG (retrieval-augmented generation)
Retrieval-augmented generation is a technique in which a system first retrieves relevant documents and then gives them to a language model as context for writing its answer. Most AI search products use some form of it, which is how they can cite current web pages. For publishers, RAG means a page can influence answers as soon as it is retrievable, without waiting for a model to be retrained.
Retrieval
Retrieval is the step in which an AI system finds candidate documents or passages that might answer a question, from a search index, a live web search or a private knowledge base. If a page is not retrieved, nothing else about it matters for that answer. Crawler access, indexing and passages that match the meaning of the question all affect retrieval.
robots.txt
robots.txt is a plain-text file at the root of a website that tells crawlers which paths they may request, using rules grouped by user agent. It follows the Robots Exclusion Protocol, which is published as an internet standard and honored by the major search and AI operators, but it is a request rather than a security control. Build AI crawler rules with the free AI crawler robots.txt generator.
S
Schema.org
Schema.org is a shared vocabulary of types and properties for structured data, founded by the major search engines and maintained as a community project. It defines types such as Organization, Product, Article and FAQPage, and properties such as name, price and author. Structured data written in JSON-LD usually uses the schema.org vocabulary.
Search crawler
A search crawler, or AI search crawler, is a bot that fetches pages to build the index an AI search product uses to find and cite sources. OAI-SearchBot and PerplexityBot are examples. For brands that want citations, these are usually the most important AI bots to allow, because they decide what an engine can quote today.
Semantic search
Semantic search is search that matches the meaning of a query rather than only its exact words. It relies on techniques such as embeddings to find passages that answer the intent behind a question, even when they use different vocabulary. It rewards content that explains a topic clearly over content that repeats a keyword.
Sentiment
Sentiment is whether an AI answer describes your brand positively, neutrally or negatively. Tracking it shows not just whether engines mention you but how they frame you, for example as a premium choice, a budget option or a product with known problems. A rising mention rate with falling sentiment calls for a different fix than low visibility does. See mentions and sentiment.
SERP
SERP stands for search engine results page, the page a search engine shows in response to a query. Modern SERPs mix classic links with features such as ads, featured snippets, People also ask boxes, local results and AI Overviews. In AI search experiences, the generated answer often takes the place of most of the SERP.
SGE (Search Generative Experience)
SGE, short for Search Generative Experience, was Google’s experimental generative AI search, offered to users who opted in through Search Labs. It tested AI-written summaries at the top of results, and its features evolved into AI Overviews. The term is now mainly historical, though you will still see it in older articles.
Share of voice
Share of voice in AI search is the portion of brand mentions in tracked AI answers that belong to your brand, compared with the competitors you track. If your mentions hold steady while competitors gain, your share of voice falls. It is one of the clearest ways to report AI visibility to stakeholders. See competitors and share of voice.
Structured data
Structured data is information added to a page in a standardized, machine-readable format that states what the page is about and its key facts, such as a product’s price, an article’s author or a store’s opening hours. It is usually written in JSON-LD with the schema.org vocabulary. It can make pages eligible for rich results in search and removes guesswork for AI systems.
T
Token
A token is the unit of text a language model processes, often a whole word or part of one. Models have limits on how many tokens they can read and write at once, and AI services often price usage by the token. Estimate the token count of a page or prompt with the free word and token counter.
Training crawler
A training crawler is a bot that collects public web content to train or improve AI models, such as GPTBot, ClaudeBot or CCBot. Allowing one can contribute to what future models know, but it does not by itself earn a citation in a live answer. Many sites make one decision about training crawlers and a separate one about search crawlers. See which AI crawlers to allow.
U
User agent
A user agent is the identifying string a browser or bot sends with each request, usually a name and version. AI crawlers announce themselves with documented user-agent names, which robots.txt rules use to target them. Because anyone can fake a user agent, checking it against the operator’s published IP ranges is needed for reliable analysis. Look one up with the free AI bot user-agent checker.
User-triggered fetcher
A user-triggered fetcher is a bot that retrieves a specific page because a person asked an AI assistant something that needed it, such as ChatGPT-User or Perplexity-User. These requests are the closest thing to a live visit from an AI user. Operators treat them differently from automated crawling, so check each operator’s documentation for how robots.txt applies.
UTM parameters
UTM parameters are tags added to a URL, such as utm_source, that tell analytics tools where a visit came from. Some AI assistants add a source tag to the links they show, which makes their traffic easier to identify. Where no tag or referrer is passed, AI visits can be misreported as direct traffic.
V
Verified bot
A verified bot is a crawler whose identity has been confirmed, usually by checking that the request comes from an IP address in the operator’s published ranges or by validating a signature the operator documents. Verification separates real AI crawlers from scrapers that borrow a famous name. aeoora counts only verified bots in its AI crawler analytics.
Voice search
Voice search is asking a question out loud to a voice assistant, smart speaker or phone and getting a spoken answer. It was an early driver of answer engine optimization, because voice assistants usually read out a single answer. Voice interfaces for AI assistants keep the same principle alive: be the clearest single answer.
Z
Zero-click search
A zero-click search is a search that ends without the user clicking through to any website, because the answer appeared on the results page or in an AI response. AI answers can make this outcome more common for informational questions. When the click never happens, being mentioned and cited inside the answer is how a brand stays visible.
Put these terms to work
Knowing the vocabulary is the first step. The next is seeing how AI engines describe your own brand today. Request a free AI visibility report to see how AI engines answer buyer questions in your category and where you are mentioned, cited or missing.
aeoora tracks how ChatGPT, Perplexity, Gemini, Claude, Copilot and Google AI Overviews and AI Mode mention, rank and cite your brand against competitors, measures AI referral traffic and AI crawler visits, and suggests the content and site fixes that close the gaps. Start a 2-day free trial, browse the free tools, or read the docs.