Glossary
The terms, defined plainly.
Every metric CitePrism reports and every term the field uses, each defined in a paragraph you can quote.
-
AI citation
An AI citation is a source link that an AI assistant attaches to its answer to show where the information came from. It is different from a mention, which is the brand named in the answer text: a brand can be named without being cited, and its site can be cited without the brand being named.
-
AI share of voice
AI share of voice is the share of sampled AI answers that name one brand, out of the answers that name any tracked brand (that brand or a competitor it tracks), across a fixed set of category questions. It counts presence per answer, not raw mention counts. If 400 sampled answers named at least one tracked brand and 100 of them named yours, your share of voice is 25 percent.
-
AI visibility
AI visibility is the degree to which AI assistants such as ChatGPT, Claude, Gemini and Perplexity name, recommend and cite a brand when people ask them questions about its category. It is measured by sampling the assistants with a fixed set of buyer questions and is reported as visibility rate, share of voice, average rank, first-mention rate and citation share.
-
Answer engine optimization (AEO)
Answer engine optimization (AEO) is the practice of structuring content so that answer engines, including AI assistants, featured snippets and voice assistants, can lift a direct answer from it. The term predates generative assistants and is now used almost interchangeably with generative engine optimization (GEO).
-
Average rank
Average rank is the mean position of a brand among the tracked brands (the brand and its competitors) named in the sampled AI answers that mention it. An average rank of 1.4 means the brand is usually the first name in the answer; an average rank of 3.8 means it is usually mentioned late, after several competitors.
-
Citation share
Citation share is the fraction of all sources cited across sampled AI answers to a category prompt set that belong to a given site, or to a category of site such as review platforms, communities or competitors. It shows which pages the assistants rely on when they answer questions in a category, and it is the closest thing generative engine optimization has to a ranking report.
-
First-mention rate
First-mention rate is the share of sampled AI answers to a category prompt set in which a brand is the first tracked brand named (the brand or a competitor it tracks). Being first in a spoken-style answer is worth more than being present, so this metric captures the part of AI visibility that visibility rate alone misses.
-
Generative engine optimization (GEO)
Generative engine optimization (GEO) is the practice of making a brand more likely to be named, recommended and cited by AI assistants when people ask them questions in its category. SEO optimizes for a ranked list of links; GEO optimizes for a single generated answer that names a few products and cites a few sources.
-
Google AI Mode
Google AI Mode is a conversational search experience inside Google Search that answers a question with a longer AI-generated response, links to its sources, and accepts follow-up questions. Google describes it as running many related searches behind one question and composing the answer from the results, a technique it calls query fan-out.
-
Google AI Overviews
Google AI Overviews are AI-generated summaries that Google Search shows above the regular results for some searches, with links to the sources the summary draws on. Google decides per search whether to show one, so an overview appears for some questions and not for others, and what it says can change over time.
-
llms.txt
llms.txt is a proposed plain-text file, written in Markdown and served at a site's root, that gives language models a short description of the site and a curated list of links to its most useful pages. Jeremy Howard of Answer.AI proposed it at llmstxt.org in September 2024. It is a convention, not a standard that any major AI assistant has publicly committed to reading.
-
Prompt sampling
Prompt sampling, also called synthetic sampling, is the method of measuring AI visibility by running a fixed set of buyer questions against AI assistants on a schedule, many times each, and analyzing the answers for brand mentions, positions and citations. It is the only data source available, because real conversations between users and assistants are private.
-
Query fan-out
Query fan-out is the process by which an AI engine turns one user question into several search queries, runs them, and composes its answer from the combined results. It means a page competes for the queries the engine writes, which often add words the user never typed, rather than for the user's original wording.
-
Share of model
Share of model is another name for AI share of voice: of the sampled answers that name any tracked brand, the share that name yours. It counts presence per answer, not raw mention counts. Read it per model, because each model answers from different data and can favor different brands.
-
Visibility rate
Visibility rate is the share of sampled AI answers to a category prompt set that mention a brand at all, regardless of position or how many other brands were named. It is the simplest AI visibility metric and the one to read first, because it describes the brand alone rather than the brand relative to competitors.