AEO Glossary
Definitions for the terms that recur across AEO work, including the ones vendors use inconsistently.
AEO (Answer Engine Optimization)
The practice of making content the source an AI assistant reaches for when answering a question. Distinguished from SEO by competing for inclusion in a single synthesised answer rather than for position in a ranked list of links.
Agent usability
Whether an AI agent acting on a user's behalf can complete a real task on your site — finding a price, checking a feature, starting a trial. Distinct from crawlability: a crawler needs to read the page, an agent needs to accomplish something. Failures are invisible in analytics because there is no conversion and no complaint.
AI Overviews
Google's generative answer panel above conventional results. Grounded in Google's index, and controlled separately from conventional crawling by the Google-Extended user agent.
Answer engine
Any system that responds to a question with a synthesised answer rather than a list of links. ChatGPT, Claude, Gemini, Perplexity and AI Overviews are the surfaces that matter commercially today.
Chunking
Splitting a document into passages for embedding and retrieval. The reason passage-level structure matters: retrieval systems score chunks, not pages, so a page's retrievability is closer to the quality of its best individual section than to the quality of the whole.
Citation
An AI answer linking to or explicitly attributing a source. On this site we distinguish citation from mention: a brand named without a link is a mention. Vendors frequently conflate the two, which inflates reported numbers.
Citation rate
The proportion of sampled answers citing a given domain, for a stated prompt set, engine and window. Meaningless without those three qualifiers and a sample count. See our measurement protocol.
Crawler serve rate
The proportion of AI crawler requests that received a successful response containing real content. A technical metric read from server or CDN logs, and the one that gates every other AEO metric.
Extractability
How easily a specific claim can be lifted cleanly from a page. Improved by answer-first structure, self-contained sections and specific checkable facts.
Fan-out (query fan-out)
The rewriting of a user's question into several retrieval queries. The reason you compete for queries nobody typed, and why covering a concept space beats targeting a phrase.
GEO (Generative Engine Optimization)
A synonym for AEO. GEO is more common in vendor marketing, AEO among practitioners. There is no meaningful difference in what the work involves. We use AEO throughout.
Grounding
Supplying a model with retrieved source material at query time so its answer rests on documents rather than only on training data. The reason a page published last week can be cited today.
Hallucination
A model stating something false with apparent confidence. In an AEO context the commercially relevant form is a false claim about your own product — a superseded price, a feature reported as missing. Correcting these is the highest return-per-hour work in AEO.
llms.txt
A Markdown file at a domain root listing a site's most useful pages with one-line descriptions. A convenience for retrieval, not an access-control mechanism and not a ranking signal. Adoption by major consumer engines is unconfirmed. See llms.txt Explained.
Mention rate
The proportion of sampled answers naming a brand, with or without a link. Usually higher than citation rate; the gap between them indicates influence without attribution.
Prompt set
The fixed collection of buyer questions used for measurement. Must be frozen and versioned, because changing it mid-study destroys comparability — the most common cause of unusable AEO data.
RAG (Retrieval-Augmented Generation)
Architecture where a model retrieves documents at query time and composes an answer from them. The mechanism underlying most AI search, and the reason retrievability matters more than being in the training data.
Re-ranking
Scoring retrieved candidate passages to select the few used in an answer. The stage where structure and specificity decide outcomes, and where content that is retrieved but never quoted is being lost.
Schema markup
Structured data, usually JSON-LD, making facts machine-readable. In AEO its value is ambiguity removal — a price in an Offer object cannot be misread — rather than ranking influence. See Schema Markup for AI Search.
Share of voice
A brand's mentions as a proportion of all brand mentions across a prompt set. The competitive metric. Requires a fixed competitor set, since quietly changing it changes the number without any underlying change.
Structured vs unstructured citation
Whether attribution appears as a visible inline link or as a collapsed footnote marker most users never expand. Affects click-through substantially while both count as citations in most tools' reporting.
Visibility score
A composite metric, usually vendor-specific, blending several underlying measures. Convenient for executive reporting and poor for diagnosis, because it obscures which component moved. Read the underlying metrics.