# Survelium documentation

> How Survelium measures your brand's presence in AI assistant answers, and how to read every number.

Survelium regularly asks AI assistants — ChatGPT, Gemini, Claude, Perplexity, Copilot and others — the same questions
(we call them **prompts**) and records what they answer: whether they mention your brand, in which position, next to
whom, in what tone, and which sources they cite. Separately, it shows which AI crawlers read your site and how many
people arrive from AI assistants.

## In short

- **[Getting started](https://survelium.com/docs/quickstart.md)** — from a new brand to your first numbers.
- **[How we collect data](https://survelium.com/docs/data.md)** — which assistants we ask, how often, in which language, and how we read each answer.
- **[Metrics](https://survelium.com/docs/metrics.md)** — what each number in the app means: the definition, the formula, an example, and what counts or not.
- **[AI traffic](https://survelium.com/docs/ai-traffic.md)** — how we see AI crawlers and visitors sent by assistants, and how to install the integration.
- **[Configuration](https://survelium.com/docs/configuration.md)** — brands, prompts, competitors, Google Search Console and reports.
- **[Guides](https://survelium.com/docs/guides.md)** — how to choose prompts and how to grow your visibility.
- **[Account](https://survelium.com/docs/account.md)** — sign-in, team, billing.

## Rules that hold everywhere

- **The default period is 28 days**, compared with the 28 days before.
- **Failed runs do not count.** If an assistant did not answer, that run is left out of every average. An empty answer
  from Google AI Overview does count: Google showed no AI Overview, so you are not visible there.
- **Competitors only count once you mark them "Direct Competitor"** on the Competitors page.
- **"—" means "not measured"**, not zero. Zero only shows when we measured and the result really is zero.
- Numbers are computed **over all answers pooled together**, not as an average of assistants: an assistant that
  answered more often weighs more.

## For AI assistants

Every page has a Markdown version (add `.md` to the address), and the list of all pages is in [llms.txt](/llms.txt).