How to use this glossary.
AI search has produced a lot of new vocabulary, much of it describing the same few ideas. AEO, GEO and LLM SEO all mean making your company easier for AI systems to find, understand and cite. Google’s own guidance says that, whatever the label, “optimizing for generative AI search is optimizing for the search experience, and thus still SEO.”
The terms are grouped alphabetically, and each links to the article that covers it in depth. If you want the practical version of all of it for your own company, our AI search visibility service for B2B firms covers technical access, content and evidence.
Term finder
Type a word, such as “crawler” or “citation”, to narrow the list of 39 terms.
How the terms fit together.
Most of these terms describe one step in a single chain, from a crawler reading your page to a buyer clicking a cited link.
The four umbrella terms overlap heavily.
| Term | What it puts first | Where it came from |
|---|---|---|
| AI search visibility | The outcome: being found, named and cited in AI answers | Industry usage; the measurable goal behind the other three |
| AEO | Being the direct answer to a question | Industry term, predates generative AI (featured snippets, voice) |
| GEO | Being cited in answers written by generative engines | Academic paper first posted 16 November 2023 |
| LLM SEO | How large language models learn and retrieve facts about you | Industry term; overlaps with GEO |
Buyer, in Google AI Mode: “Which UK payroll software providers integrate with Xero and support 200+ employees?”
The terms in play: the question is a prompt; Google splits it into several searches (query fan-out) and answers from retrieved pages (grounding); named providers get a brand mention, linked pages a citation, and clicks become AI referral traffic.
An invented question, not a real query or answer.
A: AI crawler to answer engine optimisation.
AI crawler
An AI crawler is an automated program, run by an AI company, that fetches web pages for one of three jobs: collecting training data, building a search index, or fetching a page because a user asked.
Why it matters: blocking the wrong one can remove you from AI answers while leaving training untouched. Read more: what each AI company documents about its crawlers.
AI Mode
AI Mode is Google’s conversational search mode, which answers in writing with links and takes follow-up questions. Google introduced it as a Labs experiment on 5 March 2025, using “a custom version of Gemini 2.0”.
Why it matters: Google says AI Mode is “particularly helpful for queries where further exploration, reasoning, or complex comparisons are needed”. Read more: Google AI Mode for B2B sites.
AI Overviews
AI Overviews are AI-written summaries shown in Google’s results, with links to supporting pages. Google says they “are only shown when our systems determine that it is additive to classic Search.”
Why it matters: to be a supporting link, Google says a page must be “indexed and eligible to be shown in Google Search with a snippet”. Read more: AI Overviews and zero-click data for B2B.
AI referral traffic
AI referral traffic is visits to your website from people who clicked a link in an AI assistant’s answer. Google Analytics 4 now has an AI Assistant channel, matched when the referrer is on its list of assistants such as “ChatGPT, Gemini, Deepseek, Copilot, or Grok”.
Why it matters: it undercounts, because answer engines are “not consistently passing referrer headers”, HubSpot notes. Read more: tracking AI referrals through to pipeline.
AI search visibility
AI search visibility is how often, and how accurately, your company is found, named and cited when buyers ask AI tools questions about your market. It is the outcome; AEO, GEO and LLM SEO are names for the work that improves it.
Why it matters: when a buyer asks AI for a shortlist, a company missing from the answer is missing from the shortlist. Read more: how to check your AI search visibility.
AI visibility score
An AI visibility score is a single number, usually a percentage, for how often your brand appears in a set of AI answers. Tools calculate it differently: Peec AI divides responses mentioning you by all responses; Profound divides by responses that mention at least one brand.
Why it matters: tools give the same company different scores, so compare like with like. For a human-checked starting point, see our free AI visibility report. Read more: AI visibility scores and their formulas.
Answer engine
An answer engine is a search tool that replies with a written answer rather than a list of links. Perplexity defines it as “a tool designed to give you direct, detailed answers to your questions.”
Why it matters: the answer becomes the shop window, and only a few firms get named in it. Read more: how answer engines pick and phrase answers.
Answer engine optimisation (AEO)
Answer engine optimisation (AEO) is the work of making your content the direct answer an engine gives to a question. In practice it means clear, answer-first pages on the questions buyers really ask.
Why it matters: Google treats the work as SEO, so judge AEO proposals by the questions they target and what they measure. Read more: the B2B guide to answer engine optimisation.
B–C: brand mention to Copilot.
Brand mention
A brand mention is your company’s name appearing in an AI answer, with or without a link. Ahrefs counts one “when a brand appears at least once in an AI generated response.”
Why it matters: being named puts you on the shortlist, and many mentions come from other sites describing you. Read more: the off-site sources AI uses to recommend firms.
ChatGPT ads
ChatGPT ads are paid placements shown below a ChatGPT response to users on the Free and Go plans; Plus, Pro, Business, Enterprise and Edu accounts have no ads. OpenAI began testing them in the United States on 9 February 2026, and by 21 October 2026 they were available in 89 countries.
Why it matters: a self-serve Ads Manager beta with cost-per-click bidding followed on 5 May 2026, and OpenAI says “ads do not influence ChatGPT’s answers”: it is a paid channel, separate from organic visibility. Read more: how to advertise on ChatGPT.
ChatGPT-User
ChatGPT-User is the user agent OpenAI uses when ChatGPT visits a page because a person asked it to. OpenAI says that “because these actions are initiated by a user, robots.txt rules may not apply.”
Why it matters: in your logs, it means a real person’s question led ChatGPT to your page. Read more: finding AI bots in your server logs.
Citation
A citation is a link to a specific page shown as a source inside an AI answer. Profound defines it as “a reference or link to a specific web page, article, or resource inside an answer engine response.”
Why it matters: being cited is not being recommended; your page can be cited while a competitor gets named. Read more: the page patterns AI engines quote.
Claude-SearchBot and Claude-User
Claude-SearchBot is Anthropic’s search crawler, which “navigates the web to improve search result quality for users.” Claude-User fetches pages when people ask Claude questions; disabling it “prevents our system from retrieving your content in response to a user query.”
Why it matters: both are separate from ClaudeBot, so you can opt out of training and stay in answers. Read more: AI crawlers and robots.txt.
ClaudeBot
ClaudeBot is Anthropic’s training crawler, which collects “web content that could potentially contribute to” the training of Anthropic’s AI models. Anthropic says its bots honour robots.txt, and restricting ClaudeBot signals that a site’s future content should be excluded from its training datasets.
Why it matters: blocking it is a training decision; it does not stop Claude-SearchBot. Read more: checking which AI bots reach your site.
Copilot
Copilot is Microsoft’s AI assistant, described by Microsoft as “an AI-powered assistant designed to help you be more productive”. Bing’s AI Performance report (public preview, 10 February 2026) shows when your content is cited in Copilot.
Why it matters: few vendors show your citations directly. Read more: Bing, IndexNow and Copilot.
E–G: E-E-A-T to grounding.
E-E-A-T
E-E-A-T stands for experience, expertise, authoritativeness and trustworthiness, the qualities Google’s guidance asks creators to show. Google says “E-E-A-T itself isn’t a specific ranking factor”, and “of these aspects, trust is most important.”
Why it matters: named authors, dated evidence and a clear company footprint are signals AI engines can read too. Read more: why brands go missing from AI shortlists.
Entity
An entity is a distinct, real-world thing, such as a company, person, product or place, that a search engine recognises and stores facts about. Google’s Knowledge Graph “has millions of entries that describe real-world entities like people, places, and things.”
Why it matters: if engines cannot tell your company from a similar name, they mix up your facts. Read more: when AI describes your company wrongly.
Gemini
Gemini is Google’s family of AI models and the name of its assistant app. Custom versions of Gemini power AI Mode, and “Grounding with Google Search connects the Gemini model to real-time web content.”
Why it matters: Gemini grounds answers in Google Search, so Google SEO hygiene helps here too. Read more: Google AI Mode for B2B sites.
Generative engine optimisation (GEO)
Generative engine optimisation (GEO) is the practice of improving how often and how prominently your content is cited in answers written by generative AI engines. The term comes from a paper by Aggarwal and colleagues first posted on 16 November 2023.
Why it matters: the paper reported methods that could “boost visibility by up to 40%” on its own benchmark; treat that as a research result, not a forecast. Read more: the B2B guide to generative engine optimisation.
Google-Extended
Google-Extended is a robots.txt token that controls whether content Google crawls from your site may be used to train future Gemini models and to ground their answers. Google says it “doesn’t have a separate HTTP request user agent string” and “does not impact a site’s inclusion in Google Search nor is it used as a ranking signal.”
Why it matters: blocking it is a training and grounding choice that may also limit how Gemini apps use your pages; it is not a way out of AI Overviews. Read more: what Google documents about Google-Extended.
GPTBot
GPTBot is OpenAI’s training crawler, which collects web content that may be used to train OpenAI’s generative AI foundation models. OpenAI says disallowing it “indicates a site’s content should not be used in training.”
Why it matters: blocking GPTBot does not remove you from ChatGPT search; that is OAI-SearchBot’s job. Read more: GPTBot in your server logs.
Grounding
Grounding is basing an AI answer on specific retrieved sources, such as live search results, rather than on the model’s memory alone. Google says grounding lets Gemini “cite verifiable sources beyond its knowledge cutoff.”
Why it matters: grounded answers only use pages the engine found and could read. Read more: how ChatGPT picks its sources.
K–L: knowledge cutoff to log-file analysis.
Knowledge cutoff
A knowledge cutoff is the date after which a model’s training data stops. Anything newer is only known if the model searches the web.
Why it matters: changes to your company reach AI answers through live retrieval long before training. Read more: training data versus live retrieval.
Large language model (LLM)
A large language model (LLM) is an AI system trained on large amounts of text to predict and generate language. ChatGPT, Claude, Gemini and Copilot are products built on LLMs.
Why it matters: alone, an LLM only knows its training. Read more: what LLM SEO can and cannot control.
LLM SEO / LLM optimisation
LLM SEO, also called LLM optimisation or LLMO, is improving how large language models describe and recommend your company, both in what they learned during training and in what they retrieve live. It overlaps almost entirely with GEO.
Why it matters: you control live retrieval far more than training, so most work targets what engines search today. Read more: the LLM SEO guide.
llms.txt
llms.txt is a proposed Markdown file at /llms.txt that summarises a website for AI tools. It was proposed by Jeremy Howard on 3 September 2024 “to provide information to help agents use a website.”
Why it matters: Google says Search ignores it and it will “neither harm nor help” your visibility. Read more: do you need llms.txt?
Server logs / log-file analysis
Server logs are your web server’s record of every request it receives; in Apache’s words, the access log “records all requests processed by the server.” Log-file analysis means reading those records to see which bots visited, which pages they fetched and what status codes they got.
Why it matters: it is direct proof AI crawlers can reach you, and it catches firewalls that block them silently. Read more: AI crawler log file analysis.
O–P: OAI-SearchBot to prompt panel.
OAI-SearchBot
OAI-SearchBot is OpenAI’s crawler “used to surface websites in search results in ChatGPT’s search features.” OpenAI says sites that opt out “will not be shown in ChatGPT search answers”, except as navigational links, and that robots.txt changes take about 24 hours to apply.
Why it matters: for most B2B companies this is the OpenAI bot to allow. Read more: writing robots.txt for AI crawlers.
Perplexity
Perplexity is an AI answer engine that searches the web for each question and writes an answer with numbered sources. Its help centre says “each answer includes numbered citations linking to the original sources.”
Why it matters: visible sources make it easy to see which pages are cited for your buyers’ questions. Read more: Perplexity SEO for B2B.
PerplexityBot and Perplexity-User
PerplexityBot is Perplexity’s search crawler, “designed to surface and link websites in search results on Perplexity”; it “is not used to crawl content for AI foundation models.” Perplexity-User fetches pages for a user’s question and “generally ignores robots.txt rules.”
Why it matters: blocking PerplexityBot can keep your pages out of Perplexity’s answers. Read more: how Perplexity crawls and cites sources.
Prompt
A prompt is the question or instruction someone types into an AI tool. In tracking tools it is a saved question; Profound defines it as “a query that Profound sends to answer engines on your behalf.”
Why it matters: prompts are longer than keywords, so the wording you test changes the result. Read more: how ChatGPT reads a prompt.
Prompt panel
A prompt panel is a fixed set of real buyer questions you run through AI engines repeatedly to measure visibility over time.
Why it matters: answers vary between runs, so one screenshot proves little. Our play on building a question panel gives a template. Read more: checking your AI search visibility.
Q–R: query fan-out to robots.txt.
Query fan-out
Query fan-out is when an AI search engine turns one question into several related searches and combines the results. Google describes it as “issuing multiple related searches across subtopics and data sources” for AI Overviews and AI Mode.
Why it matters: pages get retrieved for sub-questions the buyer never typed, such as integrations or locations. Read more: query fan-out in Google AI Mode.
Retrieval-augmented generation (RAG)
Retrieval-augmented generation (RAG) is a method where an AI model first retrieves relevant documents and then writes its answer using them. It was set out in a paper by Patrick Lewis and colleagues first posted on 22 May 2020, pairing a language model with “a dense vector index of Wikipedia.”
Why it matters: AI search is RAG at web scale; retrieval decides which pages the model sees. Read more: where AI answers get their sources.
robots.txt for AI crawlers
robots.txt is the plain-text file at the root of your site that tells crawlers which paths they may fetch, and AI crawlers are addressed by their user-agent names. The standard, RFC 9309 (September 2022), states “these rules are not a form of access authorization.”
Why it matters: allow search bots and decide on training bots; do not block “all AI”. Read more: writing robots.txt for AI crawlers.
S–Z: share of voice to zero-click search.
Share of voice in AI answers
Share of voice in AI answers is your brand’s share of all brand mentions across a set of AI answers, compared with competitors. Ahrefs defines it as “a brand’s percentage share of impressions compared to other tracked brands.”
Why it matters: it shows whether you are gaining on named rivals. Read more: share of voice versus visibility score.
Structured data / schema
Structured data is code on a page, usually JSON-LD using the schema.org vocabulary, that labels what the page is about. Google calls it “a standardized format for providing information about a page and classifying the page content.”
Why it matters: Google says it “isn’t required for generative AI search” and needs no special markup, but it helps SEO and must match the visible page. Read more: schema for AI search.
Training data vs live retrieval
Training data is what a model learned before its knowledge cutoff; live retrieval is what it looks up on the web while answering. Anthropic says Claude searches “when the request depends on information that is current, changing, or outside its training data.”
Why it matters: questions about specific companies often trigger a search, so your current pages count most. Read more: the two routes into an LLM’s answer.
Zero-click search
A zero-click search is one where the searcher gets what they need on the results page and does not visit any website. Pew Research found users who saw an AI summary clicked a result in 8% of visits, against 15% without one (900 US adults, March 2025).
Why it matters: being named in the answer now does part of a click’s job. Read more: zero-click searches in 2026.
Where to start with the vocabulary.
If you remember five terms, make them OAI-SearchBot, citation, brand mention, prompt panel and AI referral traffic.
- 01 · ACCESSrobots.txt and server logs: can the search crawlers read you?
- 02 · CLARITYEntity, structured data, answer-first pages
- 03 · REPUTATIONBrand mentions on the sites engines retrieve
- 04 · MEASUREMENTPrompt panel, share of voice, AI referral traffic
llms.txt comes after these, not before.
When you are ready to turn the glossary into a plan, our AI visibility work for B2B companies combines technical access, content, evidence and measurement.
Questions people ask.
What is the difference between AEO, GEO and LLM SEO?
Mostly emphasis. AEO is about being the direct answer, GEO about being cited in AI-written answers, and LLM SEO about how language models learn and retrieve facts about you. Google treats all three as SEO.
What does AI search visibility mean?
It is how often, and how accurately, your company is found, named and cited when buyers ask AI tools such as ChatGPT, Google AI Overviews, Perplexity, Gemini and Copilot about your market.
Which AI crawlers should a B2B site allow?
Usually the search crawlers: OAI-SearchBot, Claude-SearchBot and PerplexityBot, because blocking them can keep you out of those answers. Training crawlers such as GPTBot, ClaudeBot and Google-Extended are a separate choice.
Is there a single standard AI visibility score?
No. Each tool uses its own formula, prompts and engines, so scores are not comparable across tools. Run one prompt panel over time and report the trend with AI referral traffic.
Sources: Google AI guide; Google AI features; Google AI Mode; Google crawlers; Google helpful content; Google structured data; Knowledge Graph; Gemini grounding; GA4 channels; OpenAI bots; Anthropic bots; Claude web search; Perplexity bots; Perplexity help; Perplexity answer engines; Copilot; Bing, Feb 2026; OpenAI, Feb 2026; OpenAI, May 2026; OpenAI, Oct 2026; OpenAI ads help; llmstxt.org; RFC 9309; Apache logs; GEO paper; RAG paper; Ahrefs; Profound; Peec AI; HubSpot; Pew, Jul 2025. OpenAI ads pages accessed 12 November 2026; all others 13 November 2026.