# Running an audit

> Engines, models, repeats and the options you choose when you run an audit.

Source: https://youcited.com/docs/feature-guides/running-an-audit
Section: Using Cited
Last updated: 2026-10-29
All docs: https://youcited.com/docs/index.md

## What an audit does

An audit asks your questions to AI models, records every answer and the sources each one cites, and scores how often you appear. A paid audit also scans your site, runs the [site check](/docs/feature-guides/site-check) and, if you leave it on, writes [recommendations](/docs/feature-guides/recommendations).

## Engines and models

| Engine | Default models | Notes |
|---|---|---|
| **ChatGPT** (OpenAI) | a flagship and a fast model | |
| **Gemini** (Google) | a flagship and a fast model | |
| **Claude** (Anthropic) | a flagship and a fast model | |
| **Google AI** | not included by default | AI Mode and AI Overviews, as a real Google user sees them. Add them as extra models. |

The defaults move to newer models as the engines release them. The run screen lists the current models. Turn on **Compare specific models** to pick others, such as an older model your customers still use. Extra models cost a little more; the price shows before you run.

### Why several models?

Different models give different answers. A fast model is what many people get by default in a free app; the flagship is what paying users and agents often use. Checking both shows you the range.

## Three answers per question

AI answers vary from one ask to the next, so a paid audit asks each question three times per model and reports the majority result, plus how many of the three runs cited you.

## Options

- **Depth:** Standard, or Deep (slower, more detail in the analysis).
- **Recommendations:** on by default. Turn it off if you only want the scores.

## Free vs paid

| | Free audit | Paid audit |
|---|---|---|
| Questions | 5 | 5 included, more for extra |
| Engines | ChatGPT (one model) | ChatGPT, Gemini, Claude (two models each), optional Google AI |
| Answers per question | 1 | 3 |
| Site check | AI bot access only | Full site check |
| Recommendations and fixes | No | Yes |
| Report | Web page | Web page, Markdown, PDF and share link |

## How long it takes

Most audits finish in a few minutes. Deep audits and audits with many questions or models take longer. You can close the page; the audit keeps running and the dashboard shows it when it's done.

## If something fails

If an audit fails, its credits come back to you automatically and you can run it again. If one model has an outage, the rest of the audit still finishes and that model shows "This model gave no answers".

## From an agent or the CLI

Ask your connected AI assistant to "run a Cited audit for my business". It shows you the price and asks before it spends credits.

From the terminal, `cited audit quote` shows the price and `cited audit run` starts the audit after asking. `cited audit get --wait` prints the report when it's done. See [Cited CLI](/docs/integration-guides/cited-cli).