GLOSSARY
Every term in AI search visibility, defined in one breath: AEO, GEO, mention rate, citation share, prompt sets, AI crawlers, llms.txt and the rest. Each definition is 40 to 60 words, self-contained, and written to be quoted — by you in a meeting, or by an engine in an answer.
AEO (answer engine optimization) is the practice of structuring your pages so AI-driven answer systems can extract and cite them directly. Where SEO targets a ranking in a list, AEO targets being the answer itself: direct 40–60 word responses to real questions, evidence in the text, and machine-readable facts.
An AI crawler is a bot an AI company uses to fetch web pages — for model training (GPTBot, ClaudeBot), for a search index (OAI-SearchBot, PerplexityBot), or to retrieve a page live during a user's question. Blocking the search and user-fetch bots removes you from AI answers regardless of content quality.
AI Overviews are Google's AI-generated answer blocks that appear above traditional results, synthesizing a response from indexed pages and citing a handful of sources. They reach roughly two billion monthly users, and when one appears, clicks to ordinary results drop sharply — making a citation in the overview the new position zero.
AI referral traffic is the analytics segment of visitors who click through to your site from an AI product — chatgpt.com, perplexity.ai, gemini.google.com and similar referrers. It's typically a small share of total traffic today, but the sessions convert unusually well because the AI has already done the comparison work.
AIO (AI optimization) is the loosest of the AI-search acronyms — a catch-all some agencies use for any work that improves how AI systems find, interpret or recommend a business. It has no distinct methodology of its own, and it's easily confused with AI Overviews, Google's answer product, which is unrelated.
An answer engine is a system that responds to a query with one synthesized answer instead of a ranked list of links — ChatGPT, Perplexity, Gemini, Google's AI Overviews. Answer engines name a few sources and omit everyone else, which is why visibility in them is a different problem from search ranking.
Citation share is the percentage of cited sources across your tracked prompt set that belong to your domain, measured per engine against named competitors. Where mention rate counts how often you're named in answers, citation share measures how much of the evidence behind those answers is yours. Both are reported monthly.
An entity is the real-world thing — a business, person, product or place — that knowledge systems recognize and attach facts to, independent of any keyword. AI engines recommend entities, not pages. If your name, locations and facts differ across the web, engines can't confidently confirm you exist, so they name someone else.
Extraction is an engine lifting a passage from a page to use in an answer — the property that decides whether your content is quotable. Engines retrieve at the passage level, not the page level: a direct 40–60 word answer under a clear question heading gets extracted; warm-up copy gets skipped.
A featured snippet is Google's classic 'position zero' — a passage lifted from one page and shown above the ranked results as the direct answer. It's the ancestor of today's AI answers, and the writing that won snippets (direct, self-contained, question-shaped) is the same writing AI engines extract and cite.
GEO (generative engine optimization) is the work of getting a business cited and recommended inside AI-generated answers — ChatGPT, Gemini, Perplexity, Google AI Overviews. The term comes from Princeton research showing content with statistics, quotations and citations gains 22–41% more visibility in generative answers. It's the industry's umbrella term for AI search visibility work.
Grounding is how an AI engine ties its answer to retrieved sources instead of generating purely from memory — fetching current pages and constraining the response to what they say, with citations pointing back. Grounded answers are where citations come from, and being retrievable and quotable is how you get into them.
A hallucination is an AI answer stating something false with confidence — a wrong address, a product you don't sell, a fee structure you don't have. For businesses it's a reputation surface no one reviews. You can't correct the model directly, but you can fix the sources it grounds on, which is where corrections actually happen.
A knowledge graph is a database of entities — businesses, people, products, places — and the verified facts and relationships connecting them. Google's is the best known, feeding brand panels and AI answers alike. Engines trust graph-confirmed facts more than page text, so getting your entity established there stabilizes what AI says about you.
LLMO (large language model optimization) is a vendor-coined synonym for the work otherwise called GEO: making a business visible and accurately represented in LLM-generated answers. It names no distinct methodology — the tactics underneath are identical — and survives mostly in sales copy because it sounds newer than the alternatives.
llms.txt is a proposed root-level file that gives LLM crawlers a curated, markdown-formatted map of a site's best content. Adoption ran ahead of evidence: Google confirmed in June 2026 that Search ignores it, and about 97% of published llms.txt files are never fetched by any AI crawler. We ship one anyway — cheap, honest about why.
Mention rate is the percentage of answers across your tracked prompt set in which an engine names your business — whether or not it links or cites you. It's the closest thing AI search has to a rank position, measured per engine because engines rarely agree on who to name.
A prompt set is the fixed basket of buyer-intent questions you measure AI visibility against — the same prompts, run on the same schedule, across the same engines. Holding it fixed is what turns 'are we visible in AI' from an anecdote into a trend line of mention rate and citation share.
RAG (retrieval-augmented generation) is the architecture where an AI engine searches for current documents and feeds them to the model before it writes, so answers reflect the live web rather than only training data. It's why your new page can appear in ChatGPT tomorrow — and why crawlable, quotable pages matter.
Share of voice, in AI search, is your slice of total brand presence across a set of AI answers — what percentage of the mentions or citations in your tracked prompts belong to you versus competitors. It's the tool-dashboard framing of the same ground covered by mention rate and citation share.
Structured data is machine-readable markup — schema.org vocabulary in JSON-LD — that states a page's facts explicitly: what the business is, what it offers, where, at what price, reviewed how. Engines assembling answers prefer facts they can parse over prose they must interpret, making schema the cheapest credibility you can add.
Synthesis is how an AI engine composes an answer from many sources and its own trained knowledge, rather than quoting one page. Recommendations are synthesized: the engine blends reviews, mentions and comparisons into 'the three best options'. You influence synthesis by being consistently present in the material it draws from.
A zero-click search is one that ends without a visit to any website — the searcher got what they needed from the results page or an AI answer. 68% of Google searches now end this way, which is why visibility inside answers matters even when it doesn't produce a click your analytics can count.
Bring one question your customers would ask an AI. We'll run it live on the call.