AEO glossary
Sixteen terms from AI search visibility, defined precisely — and each one paired with the term it is most often confused with, because that confusion is where the reporting errors come from.
Where a term has enough behind it to deserve its own page, it links to one. The rest live here rather than becoming pages with nothing in them.
Answer engine
A system that responds to a question with a written answer rather than a list of links. ChatGPT, Perplexity, Claude, Google AI Overviews and Microsoft Copilot are all answer engines.
Why it matters: an answer names very few sources — usually two to four brands. There is no second page to be on, so visibility here is closer to binary than ranking is.
Answer engine optimization (AEO)
The practice of making a brand easy for an answer engine to find, trust and quote. It covers what is on your own pages, how that content is structured, and what the rest of the web says about you. Full version: what is AEO.
Local trap: in the Philippines, AEO usually means Authorized Economic Operator, a Bureau of Customs accreditation. “AI search visibility” is the clearer phrase there — see AI search visibility in the Philippines.
Generative engine optimization (GEO)
A near-synonym for AEO, introduced in a 2023 research paper of the same name. GEO emphasises being cited by generative models; AEO, having grown out of SEO practice, also covers featured snippets and voice results.
Confused with: AEO — and reasonably so. In practice the work is identical. The longer comparison, including the other names for the same thing, is on GEO vs AEO.
AI Overview
Google’s generated summary above conventional search results, previously branded Search Generative Experience.
Why it matters commercially: it has no public API, so nobody can monitor it programmatically, and Search Console does not break its impressions out separately from ordinary web search. Any vendor reporting AI Overviews numbers is checking by hand or scraping — ask which.
Grounding
Retrieving live documents at answer time and writing the answer from them, rather than from the model’s training data alone. Also discussed as retrieval, or RAG.
Why it matters: it is the reason publishing a page now can affect an answer within weeks rather than waiting for a model to be retrained — and the reason crawlability still matters exactly as much as it does in search.
Citation
A source an engine credits beneath or within its answer.
Confused with: a mention, and a click. All three are different. An answer can cite your page without naming your brand, and can name your brand while citing a competitor’s page about you. Neither event is a visit.
Citation share
The percentage of cited sources, across a fixed question set, that point at your own domain. Full page, with the formula and a worked example: what is citation share.
Our own figure: 0.0% — none of the 90 sources cited across our eight questions pointed at bungad.com. Published at our score.
Mention rate
The share of analysed answers in which your brand is named at all. Computed over answers, where citation share is computed over citations.
Confused with: citation share. You can have one without the other. A brand discussed widely on review sites scores well here and badly there — named often, cited never, and with no control over how it is described.
Share of voice
Your proportion of all brand mentions across the question set, counting competitors alongside you.
The trap: it is relative, so it rises when a competitor falls. A share-of-voice chart going up while your mention rate is flat is not an improvement. Read the two together or neither.
Position
Where your brand falls in the order of brands named in an answer.
The subtlety: it is only defined when you were named, so it cannot be averaged across answers that skipped you. Our own audit reports no average position at all, because we were never named — an honest “undefined” rather than a zero.
Visibility score
A single 0–100 figure combining the measures above. Ours weights mention rate at 50%, citation share at 30% and position at 20%, and the arithmetic is written out on how we measure.
Do not compare across vendors. Weightings and question sets differ, so two vendors’ scores for the same brand are not the same measurement. A score is only meaningful against itself, on a fixed prompt set, over time.
Prompt set
The fixed list of questions a measurement runs against. Also called tracked questions.
Why it is the most important item on this page: it is the denominator of every metric. Change it and the series resets. A report that does not print its prompt set cannot be verified, because quietly swapping questions can manufacture any trend you like. Ours is stored verbatim inside every run record.
Prompt tracking
Asking a prompt set to one or more engines on a schedule and recording who is named and what is cited.
What it is not: a census. It samples what an engine says when asked your questions — commercially the number that matters, but not a count of every time you were cited to anybody. More on the options: who tracks AI citations.
Zero-click answer
An answer that satisfies the question without anyone visiting a website.
Why it is invisible: analytics cannot show a visit that never happened as a loss, because it never was one. Rankings hold, impressions may hold, and the visit simply does not occur. The only instrument that detects it is asking the engines directly.
llms.txt
A proposed plain-text file at a site’s root summarising it for language models. A community proposal, not a standard. Full page, including the honest assessment: what is llms.txt.
Status: no major engine has publicly confirmed it reads one. We publish ours and have no evidence it has changed an answer.
Entity consistency
Describing your business identically wherever it appears: the same legal name, address, category, founding date and service description on your own site, your listings and your profiles.
Why it is worth a dull afternoon: contradictions between sources give an engine a reason to trust none of them. Faced with three different descriptions of you and one consistent description of a competitor, it names the competitor.
Put the vocabulary to work
Enter your domain. We ask Perplexity three of your buyers’ questions and report the mentions, the citations and the brands named ahead of you.
3 questions on Perplexity, free. A domain checked in the last week is served from that stored run rather than asked again, and the daily free allowance resets at midnight UTC.