The AEO / GEO glossary, in plain English.
Generative-AI optimization has its own vocabulary, often left unexplained. This glossary gathers the terms that come up most often — with short, jargon-free definitions and real-world examples (tourism, local business).
Missing a term, or a definition isn't clear? Tell us.
AEO (Answer Engine Optimization)
Optimizing content so it becomes the direct answer given by an answer engine (ChatGPT, Perplexity, Google AI Overviews) rather than just one result among ten blue links. In practice: structuring information so an AI can extract, understand and cite it unambiguously — structured schema, short verifiable answers, up-to-date data.
GEO (Generative Engine Optimization)
A neighboring term to AEO, sometimes used interchangeably. GEO emphasizes optimizing for generative engines broadly (not just direct answers to a question, but any AI-generated content that mentions or relies on your site — summaries, recommendations, generated comparisons).
AI citation
When an AI explicitly names your business in its answer to a user's question — as opposed to an AI that answers the question without ever mentioning you, even if your site exists and contains the information. This is the baseline metric for any AI-visibility tool (Livada Trace, Mirror).
AI share of voice
Across a representative set of queries for your industry, the proportion of AI answers where you're cited relative to your competitors. A 0% share of voice doesn't mean your site is technically broken — often the site is readable by AI but simply never gets chosen as a source, for lack of enough signals (reviews, structure, multi-platform presence).
Brand hallucination
When a generative AI states a false fact about your business — an invented price, an activity that isn't yours, confusion with an unrelated namesake or competitor. Unlike being simply absent from an answer, a hallucination is actively misleading to the person asking. Only verifiable by actually asking multiple AIs the question yourself — never by assuming it.
llms.txt
A text file at a site's root (like robots.txt), designed to give AI systems a structured, up-to-date summary of what the site offers — instead of leaving them to guess from raw HTML. It isn't (yet) a confirmed Google ranking signal, but it is a clarity signal for the AI systems that read it. Full guide: llms.txt explained →
Structured data (Schema.org)
Hidden markup in a page's code (most often JSON-LD) that explicitly describes what an entity is — a hotel, a price, a frequently asked question, a business — in a standardized vocabulary that Google and AI systems understand unambiguously. A page can read perfectly to a human and still be misunderstood by an AI if this data is missing.
Answer engine
A service that answers a question directly with a synthesized response (ChatGPT, Perplexity, Google AI Overviews, Copilot) rather than returning a list of links to explore yourself. The difference from a classic search engine isn't cosmetic: the user often never clicks through to any site — they read the answer and stop there, which is exactly why being the cited source in that answer matters.
Prompt fan-out
The mechanism by which an AI breaks a user's question into several sub-search-queries before generating its answer (for example, "best campsite in Provence" might trigger separate searches on reviews, pricing, amenities, location). Being visible for ONE phrasing isn't enough — the content needs to answer the full set of likely sub-questions.
Agent readiness
A site's ability to be not just read and cited by a conversational AI, but also navigated and used by an autonomous AI agent completing a task for the user (finding a specific piece of information, filling a form, checking availability). It's a layer beyond simple citation — see also Livada AgentRadar, which actually runs an agent through the site to verify whether it finds the right answer.
AI visibility score
A numeric score (often out of 100) summarizing how well a site is readable, understood, and cited by AI systems. Useful for tracking progress over time, but a score alone doesn't say EVERYTHING — two sites at 100/100 technically can have very different shares of voice depending on their competitors and region. Cross-check it against real citation tests, don't take it in isolation.
RAG (Retrieval-Augmented Generation)
The technique behind most current web-connected AI answers: the AI retrieves relevant pages in real time, then generates its answer grounded in that retrieved content rather than relying solely on its training memory. This is exactly why up-to-date, easily retrievable content matters more, in this context, than older content that merely ranks well on Google.
See where you actually stand
These definitions matter most against your own site. Test your AI visibility for free.
Test my AI visibility — free, 30s