Short, working definitions for the vocabulary used in generative engine optimization. Each entry explains what the term means in practice, not just what the acronym expands to.
GEO — generative engine optimization
The practice of making a brand discoverable, quotable and accurately described in AI-generated answers. It covers crawler access, content structure, corroboration and measurement. GEO is a superset of tactics, not a single technique.
AEO — answer engine optimization
Often used interchangeably with GEO. In practice, AEO emphasizes winning featured answers and direct-response formats; GEO emphasizes presence across generated, multi-source answers. The overlap is large and the distinction is mostly vendor vocabulary.
AI visibility
Whether your brand appears in AI answers for the questions your buyers ask. It is usually reported as a rate — the share of tested prompts where you are mentioned.
Mention rate
The percentage of sampled prompts where your brand is named. A mention is any appearance of the brand name in the answer text, regardless of whether the answer recommends you.
Citation rate
The percentage of sampled prompts where your domain is used as a source and linked. Citation is stricter than mention: a brand can be named without being cited, and cited pages drive traffic while bare mentions do not.
Share of voice (SoV)
Your brand mentions divided by all brand mentions in the category, across the same prompt set. Requires an explicit competitor list and a stable prompt set to be comparable.
Prompt set / prompt panel
The fixed list of questions sampled on a schedule. A prompt set is the measuring instrument; changing it invalidates trend comparisons unless the change is versioned.
AI crawler / AI bot
A program operated by an AI provider that fetches pages for retrieval or training. Common user agents: GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot, Google-Extended, CCBot, Bingbot, Applebot-Extended.
GPTBot
OpenAI's crawler for training and index building. Blocking it removes your site from that evidence set. OAI-SearchBot handles search-index crawling and ChatGPT-User handles user-initiated fetches; the three have different roles and are often configured separately.
Google-Extended
A robots.txt token that controls whether Google's Gemini and Vertex AI products may use your content, separate from Google Search indexing. Allowing it is a decision about AI visibility versus content control, not an SEO setting.
CCBot
Common Crawl's crawler. Its corpus feeds many downstream datasets, so blocking it can affect models you do not directly interact with.
llms.txt
A community-proposed Markdown file at /llms.txt summarizing a site for language models: brand description, product facts, key URLs. Not a formal standard and not a confirmed ranking factor; useful as a canonical fact sheet and crawl guide.
llms-full.txt
An extended variant containing fuller page text rather than only summaries. Adoption is lower than llms.txt.
AI Overviews
Google's generated summary shown above some search results, with citations to source pages. It is one AI surface among several; ranking well organically improves the odds of being cited there.
Grounding / retrieval
The mechanism by which a model pulls external sources into an answer. Without retrieval, an answer relies only on parametric memory, which is where stale and wrong brand facts usually originate.
Parametric memory
Facts a model absorbed during training. It cannot be updated by editing your site; the only path is better corroborated public information and retrieval-based answers.
Hallucination
A confident but incorrect statement. In GEO the relevant form is brand hallucination: wrong pricing, wrong product scope, wrong headquarters.
Claim accuracy / fact verification
Checking the factual statements AI makes about your brand against your own published facts. Distinct from mention counting — it catches "AI says we charge $99" errors.
Sentiment
Whether an AI answer describes your brand positively, neutrally or negatively. Useful as a trend; unreliable as a precise score because generated language is non-deterministic.
AI referral traffic
Visits to your site where the referrer is an AI platform. Measured in analytics; a lower bound because many AI-originated visits arrive without a referrer.
Agent-ready content
Content structured so autonomous agents can parse and act on it: stable URLs, clean HTML, explicit facts, machine-readable summaries.
Schema markup / structured data
JSON-LD annotations that describe entities and relationships to machines. It exposes facts you already publish; it does not create new ones.
Non-determinism
The property that the same prompt can produce different answers across runs. It is the main reason GEO measurement needs repeated sampling rather than single tests.