Glossary

The words behind AI visibility, defined.

21 terms, each written as a short passage that stands on its own, and each linked to the page on this site where the thing is measured or explained. Nothing here claims a number we have not published elsewhere.

GEO (Generative Engine Optimization)

The practice of making a brand citable by AI answer engines. Where classic SEO competes for a ranked link, GEO competes to be the source an AI-written answer quotes. It is measured by asking real buyer questions to the engines and recording whether the brand is named or linked in the reply.

What is GEO?

AEO (Answer Engine Optimization)

The same goal as GEO, framed around answer engines specifically: systems such as Perplexity, ChatGPT search and Google AI Overviews that return one synthesized answer instead of a list of results. In practice the two terms describe the same work and the same measurement.

What is AEO?

AI answer engine

Any product that answers a question with generated text and, usually, a set of cited sources: Google AI Overviews and AI Mode, ChatGPT, Perplexity, Gemini, Claude, Grok, Copilot, DeepSeek. BeCited measures up to seven of these on paid plans and two on the free report.

Methodology

AI Overviews

Google's generated summary shown above classic results for many queries, with links to the pages it drew from. Which pages get picked is not the same as which pages rank first; Google's own guidance says it favors content that is helpful, original and clearly attributed.

How do AI Overviews pick their sources?

AI Mode

Google's conversational search surface that answers one user question by running a set of related sub-queries in parallel and synthesizing the results. Because of that fan-out, a brand can be present for the visible question and absent for the sub-queries that actually decide the answer, or the reverse.

Query fan-out

Citation (in an AI answer)

A link or explicit source attribution inside a generated answer that points at a specific page. A citation is the strongest form of AI visibility because it sends the reader to you. It is distinct from a mention, where the brand is named in the text but no source is attached.

Mention

Mention (in an AI answer)

The brand name appearing in generated text without a link to any of its pages. A mention shows the engine knows the brand exists; a citation shows the engine used the brand's own content. Tools that count only mentions overstate visibility, so BeCited records the two separately.

Inside the CITE score

Share of voice (AI)

Of all the brands named or cited in the answers to a set of queries, the share that is yours. It is a relative number: a brand can gain share because it improved or because a competitor dropped out. It is only comparable across runs when the query set and the engines are held fixed.

Query matrix

CITE score

BeCited's 0-100 score for AI-answer visibility, built from four weighted dimensions: Citation (35%), whether engines actually cite the domain; Identity (20%), whether machines can resolve who the brand is; Trust (25%), crawlability, rendering, freshness and authorship signals; Eminence (20%), third-party presence in the sources engines quote.

Methodology

SEO Health score

BeCited's 0-100 score for classic search health, computed in the same audit as CITE: Technical (25%), On-page (20%), Content and E-E-A-T (25%), Schema (15%), Performance (15%). It exists because most AI answer engines still retrieve through a search index, so a page that cannot be indexed cannot be cited.

GEO vs SEO

Buyer query

A question a real prospect would ask an AI engine at the point of deciding: which tool for a use case, who is best for a segment, what a thing costs. Measuring visibility on brand-name queries proves little; measuring it on buyer queries, where the brand must be chosen, is what matters.

Which engine is most likely to cite you?

Query matrix

The full grid of buyer queries multiplied by engines that an audit measures, with a date on each cell. A dated matrix is what makes two audits comparable: the same queries on the same engines, so a change in the score reflects a change in the answers rather than a change in what was asked.

Methodology

Query fan-out

The technique, described in Google's own documentation of AI Mode, of answering one user question by running several related sub-queries in parallel and merging the results. It means single-query presence can understate or overstate real visibility: a measurement that asks only the visible question reads only the visible layer.

AI Mode

Repeat sampling

Asking the same query to the same engine more than once in one audit, because generated answers vary run to run. Reporting a single sample as fact overstates certainty. BeCited runs one sample on Starter and up to three on Scale, and reports measurement variance as variance rather than as a change.

Deterministic scoring

Deterministic scoring

A scoring method where the same inputs always produce the same score, so a re-run of an unchanged site returns the same number. The live engine answers still vary; determinism applies to how the collected evidence is turned into a score, which is what makes a before/after delta attributable to the work done.

Inside the CITE score

Fix attribution

Tying every fix to the audit that found it and the audit that confirmed it, so the score movement can be credited to specific work. Without attribution a rising score is just a rising score; with it, a client deck or an engineering ticket can say which change produced which delta.

Free report

llms.txt

A proposed plain-text file at the root of a site that summarizes what the site is and links its key pages, written for language models. Google states it does not use the file for Search, and no BeCited score awards points for having one; it remains a cheap way to state your own facts in your own words.

Does llms.txt work?

AI crawler

A bot that fetches pages on behalf of an AI product: GPTBot and OAI-SearchBot for OpenAI, PerplexityBot, ClaudeBot for Anthropic, Google-Extended for Gemini training. Each is named in robots.txt separately, and blocking one does not block the others. Blocking them removes the site from the answers those products give.

Do AI crawlers respect robots.txt?

AI crawlability

Whether an AI crawler can fetch a page and read its content. Two things decide it: robots.txt has to allow the bot, and the content has to be present in the server-rendered HTML, because AI crawlers do not execute JavaScript. A page that renders only on the client is invisible to them.

Do AI crawlers run JavaScript?

Entity signals

The evidence that lets a machine resolve who a brand is: a consistent name and URL, Organization or WebSite structured data, sameAs links to the brand's profiles elsewhere, and third-party pages that describe it the same way. Weak entity signals make an engine unsure whether two mentions are the same company.

Methodology

Preferred Sources (Google)

A per-reader Google Search setting: a person marks the sites they want to hear from, and Google then favors those sites for that reader in Top Stories, AI Overviews and AI Mode. Since August 2026 a site can embed a one-click button that adds it, but the button only works if the domain is already in Google's source index.

What is Google Preferred Sources?