GEO & AI search glossary
The language of AI search, defined plainly. From AEO and GEO to share of voice, citations and llms.txt: what each term means and why it matters for your brand.
AEO (Answer Engine Optimization)
The practice of making a brand more likely to be mentioned, recommended and cited in answers from AI assistants and AI search features. It focuses on the answer a person reads, not only on a ranked list of links.
Answer engine optimization →Agentic search
Search in which an AI agent runs several queries, reads sources and compares options on the user’s behalf before answering or acting. It widens the set of pages that can influence a recommendation.
AI crawler
Also: AI bot, AI user agentAn automated client operated by an AI company that fetches web pages, either to train models or to retrieve live information for an answer. Examples include GPTBot, OAI-SearchBot, ClaudeBot and PerplexityBot.
Check your AI crawler access →AI Mode
Google’s conversational search experience that answers a query with a generated response and supporting links. It is measured separately from AI Overviews and from Gemini.
Google AI surfaces →AI Overviews
Generated summaries that appear above classic results in Google Search for some queries, with links to the sources used.
AI search visibility
How often and how favourably AI systems represent a brand when people ask relevant questions: whether it is mentioned, recommended, positioned early, described accurately and supported by cited sources.
What is AI search visibility? →AI shopping visibility
Whether a brand’s products appear, and how prominently, when AI assistants recommend products or show shopping results.
AI visibility for ecommerce →Answer variance
The normal variation between answers to the same prompt over time or across runs. Reliable measurement uses repeated observations so that ordinary variance is not mistaken for a real change.
How much variance is normal? →
Brand entity
The set of verifiable facts that identify a company: its name, products, people, locations and relationships. Clear, consistent entity information makes a brand easier for AI systems to describe accurately.
Brand entity hygiene →Brand hallucination
An incorrect statement an AI system makes about a brand, such as a wrong price, feature or policy. Monitoring answers is how teams find and correct them.
LLM hallucinations about your brand →
Citation
Also: Cited sourceA link or reference an AI answer shows as a source for its content. Citations reveal which pages shape the answer and where a brand’s own or third-party evidence is missing.
AI citation gap analysis →Citation gap
A question where competitors or third parties are cited and a brand’s own evidence is not. Closing a citation gap usually means publishing or earning a stronger source for that question.
From citation gap to content brief →Competitive set
The competitors a brand chooses to compare itself with. Share of voice and other comparative metrics depend on it, so it should stay consistent between reporting periods.
Fan-out queries
Also: Query fan-outThe additional searches an AI system runs behind the scenes to answer one prompt. They explain why pages that do not rank for the original question can still be cited.
GEO (Generative Engine Optimization)
Improving how a brand is understood and represented in AI-generated answers. In practice it overlaps with AEO: better evidence, clearer entity information, crawlable content and strong third-party sources.
Generative engine optimization →Grounding
When an AI system bases its answer on retrieved sources rather than on its training data alone. Grounded answers are the ones where citations and fresh content matter most.
Large language model (LLM)
The type of AI model behind assistants such as ChatGPT, Gemini and Claude. It generates text from patterns learned in training and, in search features, from retrieved sources.
llms.txt
A proposed plain-text file at the root of a website that summarises its most useful pages for AI systems. Support varies by provider, so it complements rather than replaces a crawlable site and sitemap.
Check your site for llms.txt →
Mention
Any appearance of a brand’s name in an AI answer, whether or not it is recommended. Mention rate is the share of answers that include the brand.
Owned citation rate
The share of citations in a set of answers that point to a brand’s own domains.
Persona
A defined buyer profile used to phrase prompts and context realistically, for example a small agency owner or an enterprise IT buyer. Answers can change between personas.
Position
Also: Answer positionWhere a brand appears in an answer that lists several options. Earlier positions are usually more visible to the reader.
Prompt
The question or instruction given to an AI system. In AI visibility work, prompts are the buyer questions a team tracks repeatedly.
Prompt portfolio
A deliberately chosen set of prompts covering category, comparison, problem and decision-stage questions, with market, language and persona context. It is the scope behind every visibility number.
Build a prompt portfolio →
Recommendation
An answer that does more than mention a brand: it suggests the brand as a good choice for the user’s need. Recommendation rate is the share of eligible mentions classified as recommendations.
Retrieval
The step in which an AI system searches for and selects documents before writing an answer. Pages that are not retrieved cannot be cited.
robots.txt
A text file at a website’s root that tells crawlers which paths they may fetch. Blocking an AI crawler here can keep a site out of that provider’s answers or training.
Test your robots.txt for AI crawlers →
Sentiment
Whether an AI answer describes a brand positively, neutrally or negatively. It is read alongside mentions and recommendations, never on its own.
Visibility score
Also: AI visibilityThe share of completed answers in which a brand appears, for a defined set of prompts, platforms, market and dates. It is only comparable between periods with the same scope.
How to measure AI search visibility →
Zero-click search
A search where the user gets what they need from the results page or AI answer without visiting a website. It makes being named in the answer more important than the click.
Zero-shot answer
An answer generated from a model’s training alone, without retrieving sources. These answers reflect what the model already believes about a brand and change only when the model is updated.
Put this into practice for your brand.
Digraph tracks how ChatGPT, Gemini, Perplexity and 6 other AI systems describe your brand, finds what to fix and plans the next move with our team.