# What each number in Virent is, and what it is not. > The data model behind Virent - brands, prompts, sampled AI answers, citations, crawler requests, human sessions and attribution - and the limits of each. Canonical: https://stage.virent.app/docs/get-started/concepts Last updated: 2026-10-19 Virent keeps every metric attached to the evidence behind it. Read this page once and the rest of the dashboard will make sense. ## Three product areas, three kinds of evidence | Area | Evidence | Needs a website | Needs an install | | --- | --- | --- | --- | | Visibility | AI answers to the prompts you monitor, collected through each engine's API | No | No | | Analytics | Browser events from people, and server-side requests from AI crawlers | Yes | Yes, per collection path | | Content Agent | Opportunities, research, drafts, reviews and publications for your articles | Yes | GitHub access to publish | Each path is independent. A site can run Visibility alone, add Human Analytics for traffic, and add AI Crawler Analytics later. ## Workspace, brand and website ### Workspace Your team's account: members, plan, billing and everything monitored. Its URL is `virent.app/`. ### Brand What Visibility monitors: a company, a person or creator, a local business or anything else. A brand has a name, other names it goes by ("Also known as") and sources it owns. ### Website The domain of a brand, when it has one. Analytics, crawler data and Content Agent belong to a website and need one. A brand without a website is tracked by name. ### Competitors Other brands you monitor in the same answers, so mentions and citations have something to compare against. ## Visibility terms ### Prompt A question people ask AI assistants, such as "best invoicing tool for freelancers". Tracked prompts run once a day on every enabled engine. ### Answer One engine's reply to one prompt on one day. Answers are sampled observations: the same question can get a different answer tomorrow, or for another person. ### Mention An answer names the brand or one of its other names. Mention rate is the share of answers that name the brand. ### Citation An answer links to a source. A citation counts for your brand when it points to your website or a source you own, such as your LinkedIn or YouTube profile. ### Share of voice Your mentions as a share of all mentions of you and your competitors in the same answers. ### Position Where the brand appears in an answer, when an answer has an observable order (a list, for example). Many answers have none. Answers come from each engine's API with a model Virent selects, not from the consumer apps people use. The provider and model are shown with every answer. ## Analytics terms ### Pageview, session and visitor A session ends after 30 minutes without activity. Visitors are visitor-days: each browser counts once per day. The browser script keeps its IDs in the site's own storage and sets no cookies. ### Event, goal and funnel Events are actions you send with the SDK. A goal is a named conversion. A funnel is an ordered list of pageviews, events or goals. ### AI referral A visit that arrives from an AI assistant, such as chatgpt.com or perplexity.ai, or with a matching UTM source. ### Crawler request A request from an AI crawler to your site, captured on the server. It is a crawl, not a citation and not a person. ## How confident a conclusion is Virent keeps signals that are directly linked separate from signals that only line up in time. A page that was crawled, then cited, then visited is a connected story; it is not proof that one caused the other. The same applies to articles published with Content Agent: changes after publication are associations. ## What Virent does not claim - It does not measure real AI search volume or every conversation people have with an assistant. - It does not guarantee mentions, citations, crawls or referral traffic. - A training crawl does not establish that a page was used for training. - Over-limit usage never drops data; see [Plans and billing](/docs/workspace/plans-and-billing).