Reference
The GEO Glossary.
Every term in generative search, defined in plain English, and marked up so the engines can lift the definitions cleanly.
AI crawler
- A bot that fetches web content for AI systems, GPTBot and OAI-SearchBot (OpenAI), PerplexityBot, ClaudeBot, and Google-Extended. If robots.txt blocks them, you are invisible to the engines they feed.
AI Mode
- Google's fully conversational search experience, built on the same index and models as AI Overviews but answering multi-step questions with follow-ups instead of a single summary box.
AI Overviews
- The AI-generated summary Google places above traditional results, assembled from pages Google already ranks and cited with links, currently the highest-volume AI answer surface.
AI referral traffic
- Visits that arrive from an AI engine's cited link (chatgpt.com, perplexity.ai, gemini). Undercounted in analytics because many AI sessions arrive with no referrer and land in Direct.
AI visibility
- How often and how favorably AI engines mention, cite, or recommend your brand across the prompts your buyers actually ask.
AI visibility audit
- A structured measurement of your current AI visibility: which buyer prompts you appear in, per engine, versus competitors, with the causes of each gap traced.
AI agent
- Software that uses an AI model to act on a user's behalf, browsing, comparing, and transacting. Agents read machine-readable site signals, which is the practical case for files like llms.txt.
Answer capsule
- A 40–60 word, self-contained answer placed immediately under a heading. Most AI-cited blog posts open with a concise answer capsule like this.
Answer engine
- Any system that responds to a query with a direct answer rather than a list of links, AI chatbots, AI Overviews, and their featured-snippet ancestors.
Answer Engine Optimization (AEO)
- The trust layer of GEO: making a page eligible to be used as an answer through clarity, structure, and credibility signals.
Brand mention
- Any reference to your brand on a third-party surface, linked or not. Mentions correlate with AI-answer visibility roughly three times more strongly than backlinks.
Citation
- An explicit reference to your brand or page inside a generated answer, the core unit of GEO success.
Citation cliff
- The steep drop in citation odds as content ages: roughly half of AI citations go to pages published or updated within the last quarter.
Citation engineering
- The practice of earning AI citations deliberately: answer-ready content, original research, and placement on the third-party sources engines trust.
Citation rate
- The percentage of relevant AI answers that cite your brand or domain, tracked per engine against a fixed prompt panel.
Corroboration
- Agreement about your brand across independent sources. Engines recommend brands that reviews, communities, and editorial coverage describe consistently, not brands that only describe themselves.
Crawl access
- Whether AI bots can actually fetch and render your pages: robots permissions, server-rendered content, 200 status codes, and clean canonicals.
Digital PR
- Earning editorial coverage and mentions through newsworthy assets, original data above all. In GEO it does double duty: links for rankings, corroboration for citations.
E-E-A-T
- Experience, Expertise, Authoritativeness, Trust, Google's quality framing. In GEO terms: named authors, real credentials, first-party evidence, and consistent entity signals.
Embedding
- A numeric representation of text meaning that lets AI systems retrieve content by semantic similarity rather than keyword match.
Entity
- A distinct, machine-resolvable concept (a company, product, person). Entity clarity helps engines know who you are and trust you.
Entity graph
- The web of records that establishes your entity across sources, Wikidata, Crunchbase, LinkedIn, schema sameAs links, all agreeing on the same facts.
Fact density
- How many verifiable, sourced facts a page contains. A statistic every ~150–200 words materially increases citation odds.
Fan-out queries
- The follow-up questions an AI engine generates internally to research a complex prompt. Content that answers the follow-ups gets cited across the whole conversation.
Generative engine
- An AI system that composes answers rather than returning links: ChatGPT, Perplexity, Gemini, Claude, Copilot, and Google's AI surfaces.
Generative Engine Optimization (GEO)
- The practice of structuring content and technical infrastructure so AI engines cite and recommend your brand in their generated answers.
GEO tool
- Software that measures AI visibility, citation tracking, share of voice, prompt monitoring (Profound, Peec, Otterly). Tools measure the gap; closing it is execution work.
Grounding
- Anchoring a model's answer in retrieved, verifiable sources instead of parametric memory, the mechanism that makes citations (and GEO) possible.
Hallucination
- A confident model statement with no basis in its sources. Brands counter it with consistent, machine-readable facts everywhere the model looks.
Knowledge cutoff
- The date after which a model's training data ends. Anything newer reaches users only through live retrieval, which is why crawlability and freshness matter.
Knowledge Graph
- Google's structured database of entities and their relationships, fed by sources like Wikidata and consistent schema, and used to resolve who your brand is.
Large language model (LLM)
- The model class behind generative engines, trained on text at scale to predict and compose language, and increasingly paired with live retrieval.
llms.txt
- A proposed plaintext file that gives AI systems a concise, curated map of your site's most important content.
NAP consistency
- Identical name, address (or domain), and profile facts across every directory and listing. Inconsistency fragments your entity and weakens corroboration.
Perplexity (of text)
- A measure of how predictable text is. Clear, declarative writing lowers perplexity and is more likely to be quoted by models.
Prompt panel
- A fixed set of 50–100 real buyer prompts run against every engine on a schedule, the instrument behind trustworthy share-of-voice numbers.
Prompt volume
- How often a question is actually asked of AI engines, the GEO analogue of keyword search volume, estimated rather than reported.
Retrieval-augmented generation (RAG)
- The architecture where a model searches live sources first and composes its answer from what it finds, the reason fresh, extractable pages get cited.
Schema markup
- Machine-readable JSON-LD describing your pages (Organization, Article, FAQ, Product). It disambiguates rather than ranks: mismatched or stale markup quietly disqualifies.
Sentiment (AI)
- How AI engines describe your brand when they mention it, favorable, neutral, or warning, tracked alongside citation rate.
Share of voice (AI)
- The share of relevant AI answers that cite you versus competitors, for a defined set of buyer prompts.
Snippet eligibility
- Whether Google may quote your page (no nosnippet, no restrictive max-snippet). If Google can't quote you, AI Overviews can't cite you.
Source overlap
- How much two engines share cited domains. ChatGPT and Perplexity overlap only about 11%, which is why per-engine playbooks exist.
Stochasticity
- The run-to-run randomness of model answers. It's why single-shot visibility checks mislead and trend lines over repeated panels are the standard.
Unlinked mention
- A brand mention with no hyperlink. Worthless in classic SEO accounting, valuable in GEO, models read text, not just links, making reclamation outreach double-purpose.
Zero-click search
- A query resolved entirely on the results or answer surface. In AI search the 'click' you compete for is the citation itself.