Glossary

GEO glossary: the terms you need, explained without jargon

This glossary covers the 34 most common terms in GEO (generative engine optimization): concepts such as grounding and query fan-out, metrics such as mention rate and share of voice, and the crawlers and technical files that decide whether AI can read your site. They are the same definitions GeoRush uses to measure a brand’s visibility.

Updated on

Core concepts

TermDefinition
GEOGenerative engine optimization: practices that get AI assistants to mention your brand, describe it accurately and cite your site. See what is GEO.
AEOAnswer engine optimization. Nearly a synonym for GEO; it started with featured snippets and voice assistants and is now used for chatbots too.
LLMLarge language model: the model that generates text, such as GPT, Gemini, Claude or Mistral. The assistant is the product built on top of it.
Generative engineA system that writes an answer instead of returning a list of links: ChatGPT, Gemini, Claude, Perplexity, Grok, DeepSeek, Le Chat, AI Overviews, AI Mode.
AI OverviewsAI-generated summaries that Google shows above the results for some searches, with links to the pages they draw on.
AI ModeGoogle Search’s conversational mode: longer generated answers, follow-up questions and links to sources.
Query fan-outA technique where the system splits one question into several related searches (subtopics, sources) and combines the results. Google documents it for AI Overviews and AI Mode.
GroundingAnchoring an answer in sources retrieved at the moment (web search, documents) rather than only in the model’s memory. It reduces errors and makes citations possible.
RAGRetrieval-augmented generation: the model retrieves relevant documents and uses them to write the answer. It is the technique behind grounding.
Knowledge cutoffThe date a model’s training data ends. Anything later, it only knows if it searches the web.
HallucinationA false statement the model presents as fact: an invented price, a feature that does not exist. That is why you should check what AI says about your brand.
PromptThe text sent to the assistant. In GEO, prompts are the questions a real buyer would ask: “best X for Y”, “alternatives to Z”.

AI visibility metrics

TermDefinition
MentionThe answer names your brand, with or without a link.
CitationThe answer links a page on your site as a source. You can be mentioned without being cited, and vice versa.
Mention rateThe percentage of analyzed answers that include your brand.
Share of voiceYour mentions divided by all brand mentions in your category across those answers.
Citation rateThe percentage of answers that cite your domain as a source.
Average positionThe average place your brand holds when AI lists options (1 = first).
SentimentThe tone the answer uses for your brand: positive, neutral or negative.
Confidence intervalThe range where a metric’s true value probably lies. With few answers it is wide. GeoRush uses a 95% Wilson interval for mention rate.
GEO ScoreGeoRush’s 0–100 index combining five weighted pillars: visibility, share of voice, citation, reputation and technical readiness.
AI referral trafficVisits that reach your site from an assistant (for example chatgpt.com or perplexity.ai). ChatGPT adds utm_source=chatgpt.com to its links.

Crawlers, files and technical data

TermDefinition
robots.txtA file at your site root that tells each crawler what it may read. It is a courtesy standard (RFC 9309), not access control. See robots.txt for AI bots.
Training crawlerA bot that collects content that may be used to train models: GPTBot (OpenAI), ClaudeBot (Anthropic), CCBot (Common Crawl), meta-externalagent (Meta).
Search crawlerA bot that indexes pages so the assistant can find and cite them: OAI-SearchBot, Claude-SearchBot, PerplexityBot, Googlebot, Bingbot.
User-triggered fetcherAn agent that visits a page because a user asked for it right then: ChatGPT-User, Claude-User, Perplexity-User. Some may not follow robots.txt.
Google-ExtendedA robots.txt token (not a separate crawler) that controls whether Google may use your content to train Gemini and for grounding in Gemini Apps and Vertex AI. It does not affect Google Search.
Applebot-ExtendedApple’s token for excluding your content from training its models. Blocking it does not remove you from Apple’s search results.
llms.txtA Markdown file proposed at llmstxt.org that summarizes your site for models and agents, with links to key pages. Use by the major assistants is not confirmed. See what is llms.txt.
nosnippetA directive asking Google not to show snippets of a page; it also keeps the page from being used as a source in AI Overviews and AI Mode.
Structured dataInformation about a page in a standard format (the schema.org vocabulary) that search engines read unambiguously: organization, product, FAQ, reviews.
JSON-LDThe format Google recommends for adding structured data to a page, inside a script tag.
E-E-A-TExperience, Expertise, Authoritativeness, Trust: criteria from Google’s quality rater guidelines for assessing content. It is a framework, not a single technical signal.
EntityA uniquely identifiable thing (a brand, person or product) that search engines and models recognize along with its attributes. A consistent entity is easier to recommend.

How to use this glossary

The terms fall into three layers worth tackling in order. First, the technical layer: search crawlers can read you (robots.txt, indexing, structured data). Second, the content and reputation layer: pages that answer directly, and a presence in the sources AI cites. Third, the measurement layer: mention rate, share of voice, citations, position and sentiment, measured over enough answers that changes are reliable.

To see these metrics applied to your brand, GeoRush calculates them across 9 engines in AI visibility, and the free check gives you a first approximation with no account.

Go from definitions to your own data

Check for free whether AI mentions your brand when your customers ask about your category.

Sources

Official documentation and primary sources checked on November 16, 2026. These policies change often: check the source before making important decisions.

Frequently asked questions

In practice, very little: both mean optimizing so that answer-generating engines mention you. AEO is the older term, tied to featured snippets and voice assistants; GEO focuses on language models.

It is when an assistant links a page on your site as a source for its answer. It differs from a mention, which is when it names your brand, with or without a link.

It is the percentage your mentions represent out of all brand mentions in your category across a set of answers. It is how you compare yourself with competitors.

It is when an assistant bases its answer on sources it retrieves at that moment, such as search results, instead of only on what it learned in training. It is what allows it to cite current pages.

No. According to Google, it is a robots.txt control token with no user agent of its own. It controls whether your content is used to train Gemini and for grounding in Gemini Apps and Vertex AI, and it does not affect Google Search.

Keep reading