Back to home

AI Search & GEO Glossary

Plain-English definitions of the terms behind Generative Engine Optimization and AI visibility. New to the field? Start with what GEO is.

AI crawler
An AI crawler is a bot that fetches web pages for AI systems — for training (e.g. GPTBot, ClaudeBot) or for live answer retrieval (e.g. PerplexityBot, OAI-SearchBot). Most AI crawlers do not execute JavaScript, so content that only appears after client-side rendering is invisible to them.
AI visibility
AI visibility is the degree to which a brand is named, described, and recommended in AI-generated answers. A brand with high AI visibility appears consistently when assistants answer questions in its category; a brand with low AI visibility is silently omitted, no matter how well it ranks in classic search.
Answer engine
An answer engine is a system that responds to a query with a single synthesized answer instead of a list of links — ChatGPT, Gemini, Claude, and Perplexity are the prominent examples. Because an answer names only a handful of brands, being excluded from it means being invisible for that query.
Answer Engine Optimization (AEO)
Answer Engine Optimization (AEO) is a near-synonym of GEO: optimizing content and brand signals so answer engines select, quote, and recommend you. Some practitioners use AEO for featured-snippet-style optimization and GEO specifically for generative AI answers; in practice the techniques overlap heavily.
Answer-shaped content
Answer-shaped content is a page structured to directly answer one specific question, with the answer stated plainly in the first sentence and supporting detail after it. Answer engines preferentially retrieve and quote this format because it can be lifted into a response with minimal rewriting.
Brand hallucination
A brand hallucination is an AI assistant stating something false about a brand — wrong pricing, discontinued products, misattributed features, or confusing it with a similarly named company. Regular monitoring across platforms is the only way to catch hallucinations before customers act on them.
Brand mention
A brand mention is any appearance of a brand name in the text of an AI answer, whether as a recommendation, a comparison, or a passing reference. Mentions in the answer body generally matter more than footnote citations, because they are what the user actually reads.
Citation (AI answer)
A citation is a source link an AI assistant attaches to its answer to show where information came from. Being cited means the assistant retrieved and used your page; tracking citations reveals which of your pages — and which third-party pages about you — AI engines actually trust.
Competitor tracking (AI)
AI competitor tracking records which other brands appear in the same AI answers as yours — who is named alongside you, ahead of you, or instead of you. It turns "we're not in the answer" into the actionable "these three brands are, and here's where they're cited from."
Consensus signal
A consensus signal is agreement across many independent sources about what a brand is and how good it is. AI models weight repeated, consistent descriptions across reviews, directories, and editorial coverage far more heavily than a brand's own claims about itself.
E-E-A-T
E-E-A-T stands for Experience, Expertise, Authoritativeness, and Trustworthiness — the qualities Google's quality guidelines reward and that AI systems approximate through consensus and citation patterns. Transparent methodology pages, named authors, and third-party corroboration all feed it.
Entity (knowledge graph)
An entity is a uniquely identifiable thing — a company, product, or person — that search engines and AI models track across the web, independent of exact wording. Consistent naming, structured data, and profile links (sameAs) help systems consolidate your brand into one strong entity instead of scattered fragments.
Evoked set
The evoked set is the shortlist of brands an AI assistant (or a human) actually considers when answering a category question. GEO is largely the work of getting into — and staying in — the evoked set that models associate with your category.
Generative Engine Optimization (GEO)
Generative Engine Optimization (GEO) is the practice of improving how often, and how favorably, a brand appears inside AI-generated answers from assistants like ChatGPT, Gemini, Claude, and Perplexity. Where traditional SEO competes for a ranked position on a results page, GEO competes for a mention inside the answer itself.
GEO audit
A GEO audit is a structured assessment of how AI assistants currently treat a brand: visibility per platform, rank within answers, share of voice against competitors, citation sources, and sentiment — plus the on-site and off-site gaps causing weak spots. It is the day-zero baseline every GEO program should start from.
Grounding
Grounding is when an AI assistant bases its answer on retrieved, up-to-date sources (like live web results) rather than only its training data. Grounded answers can change as the web changes — which is why brands can improve their AI visibility much faster on grounded engines like Perplexity than on pure recall.
Large Language Model (LLM)
A large language model is the neural network underlying AI assistants, trained on vast text corpora to generate language. Its parameters encode what it "knows" about brands from training; retrieval (RAG) supplements that knowledge with live sources at answer time.
llms.txt
llms.txt is a proposed convention — a plain-markdown file at the root of a website that gives AI systems a concise, structured summary of who the company is, what it offers, and where its key pages are. It is the AI-crawler counterpart of robots.txt plus a sitemap, optimized for being read rather than indexed.
Prompt intent
Prompt intent is the underlying goal behind a question asked to an AI assistant — comparing options, seeking alternatives, validating a choice, or asking for a recommendation. Brands can rank well for one intent and be absent for another, so visibility must be measured across the intent spectrum.
Retrieval-Augmented Generation (RAG)
Retrieval-Augmented Generation (RAG) is the technique behind grounded answers: the system first retrieves relevant documents, then generates a response conditioned on them. For brands, RAG means your (and your reviewers') live web content directly shapes what AI assistants say about you.
Sentiment (in AI answers)
Sentiment in AI-answer tracking is whether an assistant's description of a brand is positive, neutral, or negative. Being mentioned is not enough — an answer that names your brand while recommending a competitor, or names it with caveats, carries very different business value.
Server-side rendering (SSR)
Server-side rendering means a page's content is present in the HTML the server returns, rather than being assembled in the browser by JavaScript. Because most AI crawlers and the first pass of search crawlers do not run JavaScript, content that is not server-rendered effectively does not exist for them.
Share of voice (SoV)
In AI search, share of voice is the percentage of relevant AI answers that mention your brand relative to all brands mentioned for the same prompts. It shows who "owns" a category in the eyes of AI assistants — and who is being recommended instead of you.
Source rank (citation rank)
Source rank is the position of a page in the citation or source list an AI assistant attaches to its answer — S#1 is the first cited source. It measures whether your content is being retrieved and relied on at answer time, independent of whether your brand is named in the answer text.
Structured data (schema markup)
Structured data is machine-readable JSON-LD markup (schema.org vocabulary — Organization, Product, FAQPage, BlogPosting) embedded in a page to state facts unambiguously. It removes guesswork for search engines and AI systems alike: instead of inferring your prices or FAQ answers from prose, they can read them as data.
Text rank
Text rank is the position at which a brand is first mentioned within the body of an AI answer — T#1 means it was named first, T#4 means three other brands were named before it. Earlier text rank correlates with stronger recommendation weight in the answer.
Tracked prompt
A tracked prompt is a realistic buyer question ("what is the best CRM for a small agency?") that is run repeatedly against AI platforms to measure whether a brand appears in the answers. A representative prompt set across intents — comparison, recommendation, alternatives — is the foundation of AI visibility measurement.
Training-data recall
Training-data recall is what an AI model "remembers" about a brand from its pre-training corpus, without consulting the live web. It shifts slowly — only as models are retrained — which is why consistent third-party coverage over time matters more for recall than any single new page.
Visibility score
A visibility score quantifies how often a brand appears when AI assistants answer a representative set of category questions, typically expressed from 0 to 100 per platform. It is computed by running many prompts against each assistant and measuring the share of answers that mention the brand.

Want to see these metrics for your own brand? Get started with geopsy.