# GEO audits

Your dashboards tell you _what's happening_ — where you show up, who's ahead, what AI says about you. A **GEO audit** tells you _what to do about it_. It's an on-demand, AI-driven analysis of a single tracked prompt that turns your visibility data into a scored diagnosis, a prioritised list of fixes, and a sequenced plan you can hand to your team. It's the flagship feature of the **Take action** workflow. To run one, see [Running an audit](/docs/running-an-audit).

> GEO audits are in **beta**. The shape of the report may change as we refine it.

## What an audit analyzes

An audit focuses on **one prompt** at a time. It reads everything Voxoria has captured for that prompt — the answers your engines produced, the sources they cited and retrieved, the competitors they named, and how your own brand appears across them — and it reads your brand's own pages to judge them against what the engines reward.

The AI writes the _judgement_: the verdict, the scores, the gaps, the recommendations. The _numbers_ behind it — your share of voice, where you rank, which answers mention you — are computed directly from your tracked data, not generated by the model. So the analysis is interpretive, but the facts it rests on aren't invented.

Each audit costs **100 credits**. See [Running an audit](/docs/running-an-audit) for how credits work.

## The verdict

Every audit opens with a headline read on where you stand for this prompt:

- **Overall standing** — one of **absent**, **emerging**, **present**, or **leading**, depending on how strongly AI engines feature you.
- **On-topic** — whether the prompt actually fits your product (**on-topic**, **partially on-topic**, or **off-topic**). A prompt the engines answer with a different category isn't a fair fight.
- **Winnable** — flagged when the evidence says you can realistically move the needle here.

Alongside it, a strip of headline numbers: your **weighted rubric score** (with a target to aim for), **share of voice** for the prompt, **tracked mentions** across engines, and whether your **brand is present** at all.

## What's in the report

The report follows the same arc as the work it's recommending — **diagnose**, **prioritise**, **execute** — across a set of tabs.

### Overview

The executive summary. A plain-language read of where you stand, **the opportunity** (how far your weighted score could move by closing the gaps the audit found), and your **top three priorities** pulled from the full action plan.

### Diagnosis

The deep diagnostic. A **scored rubric** grades your prompt across several dimensions tailored to your brand and category. Each dimension carries a **current score and a target** (on a 0–10 scale) and a **weight**, and the rubric is ordered by the biggest _lever_ first — the gap that, closed, moves your score the most. Expand any dimension to see what's missing, what engines reward for it, and the evidence behind the score. Below the rubric, **blocking gaps** rank the specific barriers keeping you from being recommended.

### Competition

The **competitive field** — the brands engines named for this prompt, how often each appears, and where you sit among them — plus the **category lanes** the engines pull from and whether you're present in each.

### Channels

A per-engine breakdown: how many runs each engine produced, what it cited, the domains it leans on, and whether you appeared. Supporting **evidence** snippets show the actual quotes and citations behind the findings.

### Action plan

The roadmap. An **effort × impact matrix** shows where to start — the high-impact, low-effort moves are the sweet spot. Below it, the full plan is grouped into **phases**, each with a goal and a rough timeline. Every action carries a **priority**, an **expected impact**, an **effort** rating, the **rubric dimension it fixes**, and a concrete **deliverable** with a **win condition** and a **hand-off spec** detailed enough to pass straight to whoever does the work. A closing **roadmap** lays out the phases and a target for your next review.

## Where to go next

Ready to run one? See [Running an audit](/docs/running-an-audit). To understand the metrics an audit builds on, see the [Metrics reference](/docs/metrics).