# Running an audit

A [GEO audit](/docs/audits) analyses one of your tracked prompts and produces a full action plan. This page covers how to start one, what it costs, and what to do with the result.

## Before you start

Two things need to be in place:

- **An own brand.** Your project must have its own brand set — the audit measures _you_ against the field, so it can't run without one. This is created during onboarding; see the [Quickstart](/docs/quickstart).
- **At least one active prompt.** Audits run against a single prompt, chosen from your **active** prompts. See [Prompts](/docs/prompts) for the active/inactive model.

Each audit costs **100 credits**. Credits are the prepaid units you spend on actions, and today audits are the only thing that uses them — your plan includes a monthly allowance, and you can top up if you run out.

## Create an audit

1. Open the **Audits** page from the sidebar.
2. Click **Create audit**.
3. Choose the **prompt** to analyse and give the audit a **name** (e.g. "Q3 visibility review").
4. Click **Create**.

The audit is queued and you're taken straight to its page. Creating an audit needs edit access to the project — if you don't see the **Create audit** button, your role is read-only.

If your account is short on credits, the dialog tells you so and offers a top-up link rather than starting a run you can't pay for.

## While it runs

An audit usually takes a few minutes. Its page shows a **running** state and refreshes on its own — there's no need to reload. The agent is reading the prompt's captured answers, checking your pages, and assembling the rubric and action plan as it goes.

You can leave the page and come back; the audit keeps running, and you'll find it at the top of the **Audits** list when it's done.

## When it's done

The finished report opens on its verdict and tabs — see [GEO audits](/docs/audits) for what each section means. Two actions sit at the top:

- **Export PDF** — produces a clean, printable version of the full report, every tab expanded, for sharing outside the app.
- **Share** — copies a link to the report. Anyone on your team with access to the project can open it.

## Re-running and progress

You can audit the same prompt again any time — just create another audit for it. Each run is **independent**: it tailors its own rubric to what the engines are saying _now_, and it costs another 100 credits. Re-running after you've shipped some of the recommended fixes is the intended way to measure progress: the new verdict and score tell you whether the work moved the needle.

## If an audit fails

Occasionally a run can't complete. The audit is marked **failed** and the 100 credits are **not charged** — the hold is released back to your balance. Just create a new audit to try again, and if failures persist, contact support.