Glossary

GEO and AI search, defined.

The words that come up when people talk about GEO and AI search, defined in plain English, with links between related terms.

A

AEOAnswer engine optimization: Answer engine optimization: getting a brand or page used in the answers that AI assistants and search features write. It means the same thing as GEO in practice. AEO is the older term, first used for featured snippets and voice assistants.; See also: GEO, Answer engine, LLMO
AI crawler: A bot an AI company uses to fetch web pages. There are three kinds. Training crawlers collect pages to train future models. Search crawlers build the index an assistant searches when it answers. User fetchers read a specific page while answering someone's question. Each can be allowed or blocked separately in robots.txt, and blocking a search crawler can remove a site from that assistant's answers.; See also: GPTBot, OAI-SearchBot, Claude-SearchBot, PerplexityBot, robots.txt
AI Mode: Google's conversational search mode, where a whole results page is an AI-written answer with links, and people can ask follow-up questions. Google reported more than a billion monthly users in July 2026. It draws on the normal Google index, so a page that can appear in Google Search can be cited in AI Mode.; See also: AI Overviews, Query fan-out, Grounding
AI Overviews: The AI-written summary at the top of many Google search results, with links to the pages it drew on. When one appears, fewer people click the ordinary results below it, but brands cited inside it keep more of their clicks than brands that aren't.; See also: AI Mode, Zero-click search, Citation
AI visibilityVisibility, Mention rate: The share of AI answers that name a brand, across a set of buyer questions. If a brand is named in 30 of 120 answers, its visibility is 25%. Because AI answers change from one run to the next, this share is a steadier measure than any single answer or rank.; See also: Share of voice, Position, AI visibility tracking
AI visibility trackingPrompt tracking, LLM tracking: Putting a fixed set of buyer questions to AI assistants on a regular schedule and recording who gets named, in what order and from which sources. Tracking over time, under the same conditions each time, is what turns noisy individual answers into a trend.; See also: AI visibility, Prompt, Share of voice
Answer engine: A search tool that replies with a written answer instead of a list of links. ChatGPT with search, Perplexity, Google's AI Mode and Gemini are all answer engines.; See also: AEO, GEO

B

Brand mention: Any page that names a brand, linked or not. In AI search, mentions on reviews, forums, videos and articles matter more than they did in classic SEO. Ahrefs found web mentions correlated with AI Overview visibility about three times as strongly as backlinks.; See also: Off-site optimization, Citation

C

ChatGPT search: ChatGPT's ability to search the web and cite pages in its answers. It switches on by itself for questions that need current information, such as product recommendations and prices. OpenAI's OAI-SearchBot crawler builds the index it searches.; See also: OAI-SearchBot, ChatGPT-User, Grounding
ChatGPT-User: The user agent ChatGPT uses when it visits a page because of something a user asked, such as reading a link they pasted. OpenAI says it isn't used to decide what appears in ChatGPT search.; See also: OAI-SearchBot, GPTBot, AI crawler
CitationSource: A link an AI answer shows to the page it drew on. Citations are the best evidence of what shaped an answer. They show which reviews, threads, articles and brand pages the assistant read, and whether those pages mention you or your competitors.; See also: Grounding, Brand mention, AI Overviews
Claude-SearchBot: Anthropic's crawler for indexing pages to improve Claude's search results. Anthropic says blocking it may reduce a site's visibility in Claude's answers. It's separate from ClaudeBot, which collects training data.; See also: Claude-User, ClaudeBot, AI crawler
Claude-User: The user agent Claude uses to fetch a page while answering a user's question. Anthropic says blocking it stops Claude retrieving the site's content for those answers.; See also: Claude-SearchBot, ClaudeBot, AI crawler
ClaudeBot: Anthropic's crawler for collecting public web content that may be used to train its models. Blocking it keeps a site's future content out of training data. It doesn't control whether Claude can cite the site in search answers. Claude-SearchBot and Claude-User do that.; See also: Claude-SearchBot, Claude-User, Training data

D

Deep research: A mode in ChatGPT, Gemini, Claude and Perplexity where the assistant spends several minutes running dozens of searches and writes a long report with many citations. Buyers use it for bigger decisions, such as choosing software. It reads far more sources than a normal answer, which gives niche pages more of a chance.; See also: Query fan-out, Citation

G

GEOGenerative engine optimization: Generative engine optimization: the work of getting a brand named and cited when people ask AI assistants questions. It covers the brand's own site, its presence on the sources assistants read, such as reviews, forums, video and press, and measuring the result over time. The term comes from a 2023 research paper from Princeton and IIT Delhi.; See also: AEO, LLMO, AI visibility
Google-Extended: A robots.txt token that controls whether Google may use a site's content to train Gemini models and to ground answers in the Gemini app and Vertex AI. It doesn't affect Google Search, AI Overviews or AI Mode. Blocking it can still keep a site out of the Gemini app's answers.; See also: robots.txt, Grounding, AI crawler
GPTBot: OpenAI's crawler for collecting content that may be used to train its models. Blocking it keeps a site out of future training but doesn't remove it from ChatGPT search. That's controlled by OAI-SearchBot.; See also: OAI-SearchBot, ChatGPT-User, Training data
GroundingWeb search, Live retrieval: When an AI model looks up current information, usually by searching the web, before it answers, and bases the answer on what it found. Grounded answers cite their sources. Most product and brand recommendations from assistants today are grounded, which is why the pages they read matter more than what the model learned in training.; See also: RAG, Citation, Training data

H

Hallucination: A confident statement from an AI model that's false, such as a wrong price or a feature a product doesn't have. Grounding reduces hallucinations but doesn't stop them. Many errors about brands actually come from real pages that are out of date, so fixing the source often fixes the answer.; See also: Grounding, Sentiment

K

Knowledge cutoffTraining cutoff: The date a model's training data ends. Anything after it is unknown to the model unless it searches the web. Cutoffs are usually months behind the release date, so a model's built-in picture of a market is always somewhat out of date.; See also: Training data, Grounding

L

Large language modelLLM: The kind of AI model behind ChatGPT, Claude, Gemini, Grok and others. It's trained on a very large body of text to predict and generate language. On its own it only knows its training data. Connected to search, it can read and cite current pages.; See also: Training data, Grounding, Knowledge cutoff
LLMOLLM optimization, AI SEO, AI search optimization: Large language model optimization. Another name for GEO, along with AI SEO and AI search optimization. They all mean getting a brand into AI-written answers.; See also: GEO, AEO
llms.txt: A proposed file at a site's root that lists its key pages in Markdown for AI tools. Google says its Search, including AI Overviews and AI Mode, ignores llms.txt, so it neither helps nor hurts there. It's cheap to add but shouldn't be anyone's priority.; See also: robots.txt, Structured data

O

OAI-SearchBot: OpenAI's crawler for the index behind ChatGPT search. OpenAI says sites that block it won't be shown in ChatGPT search answers. It's the most important OpenAI crawler to allow if you want to be recommended in ChatGPT.; See also: GPTBot, ChatGPT-User, ChatGPT search
Off-site optimization: GEO work that happens on other people's websites: review profiles, roundup articles, forum discussions, videos, directories and press. Because AI answers draw heavily on these sources, off-site work often decides whether a brand gets recommended.; See also: Brand mention, Citation, GEO

P

PerplexityBotPerplexity-User: Perplexity's crawler for the index it answers from. Perplexity says it isn't used to train foundation models. A second user agent, Perplexity-User, fetches pages while answering a specific question.; See also: AI crawler, robots.txt
Position: Where a brand is named in an AI answer: first, second, and so on. Readers pay more attention to the first names, so it matters. But the order changes more between runs than which brands appear, so average position over many answers is a secondary measure to visibility.; See also: AI visibility, Share of voice
Prompt: What someone types into an AI assistant. In GEO, a tracked prompt is a question a real buyer might ask, such as "best payroll software for a 20-person company in the UK", checked regularly to see who gets named.; See also: AI visibility tracking, Query fan-out

Q

Query fan-out: How AI search turns one question into several smaller searches, reads the results of each and writes one answer from them all. Google describes AI Overviews and AI Mode as running "multiple related searches across subtopics and data sources". It's why pages that don't rank for the exact question can still be cited: they rank for one of the sub-questions.; See also: Grounding, AI Overviews, AI Mode

R

RAGRetrieval-augmented generation: Retrieval-augmented generation: a system that finds relevant documents first and gives them to the model to write its answer from. AI search is RAG over the web. Grounding is the same idea from the user's side.; See also: Grounding, Large language model
robots.txt: A file at a site's root that tells crawlers which pages they may fetch. For AI search, it's where a site can allow answer crawlers such as OAI-SearchBot, Claude-SearchBot and PerplexityBot while blocking training crawlers such as GPTBot and ClaudeBot. A careless rule here is one of the most common reasons a brand is missing from AI answers.; See also: AI crawler, OAI-SearchBot, GPTBot

S

Sentiment: How an AI answer describes a brand: the reasons it gives for recommending it, and the drawbacks it mentions. These come from the reviews and discussions the assistant read, so a common complaint online tends to show up in answers.; See also: Hallucination, Citation
Share of voice: A brand's share of all company mentions in a set of AI answers. If the answers name 240 companies in total and 30 of those mentions are the brand, its share of voice is 12.5%. It shows how much of the conversation a brand gets compared with its competitors.; See also: AI visibility, Position
Structured dataSchema markup: Code on a page, usually schema.org in JSON-LD, that labels facts like prices, ratings and organization details for machines. Google says it isn't required for its AI features, but it still helps ordinary search results and costs little to add.; See also: llms.txt

T

Training data: The text a model learned from before release. It shapes which brands a model knows about and how it describes a category, but it's fixed at the knowledge cutoff. Changing it takes a new model. Changing what the model reads when it searches can happen in weeks.; See also: Knowledge cutoff, Grounding, GPTBot

Z

Zero-click search: A search that ends without a click to any website, because the answer was on the results page. AI answers make it more common. Pew found people clicked a result on 8% of Google visits with an AI summary, against 15% without. In zero-click search, being named in the answer is the visit.; See also: AI Overviews, AI Mode

See where you stand in AI answers.

Bring your brand, or one client, and the questions its buyers ask. On a 30-minute call we'll show you what six AI assistants say, and what we'd fix first.

Book a demo See pricing