> ## Documentation Index
> Fetch the complete documentation index at: https://chatobserver.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# How do I add and run my first AI monitoring prompt?

> Create your first prompt in chatobserver, select an AI platform and region, and open the collected answer.

Open **Prompts**, choose **New prompt**, and enter the question you want to monitor. Saving a prompt creates its monitoring configuration. Running it collects an answer.

<Frame caption="Choose the question, platform, region, and schedule before creating the prompt.">
  <img src="https://mintcdn.com/chatobserverltd/5Yfvp5xLWzwKiCcC/images/documentation/add-prompt.webp?fit=max&auto=format&n=5Yfvp5xLWzwKiCcC&q=85&s=f62059f8ec17b0d02d0aea26cc455647" alt="Add a prompt dialog with a question, ChatGPT selected, a region and a schedule" width="512" height="482" loading="lazy" decoding="async" data-path="images/documentation/add-prompt.webp" />
</Frame>

## Create the prompt

1. Enter a complete question using the language your audience would use.
2. Choose the platforms and regions. ChatGPT is selected by default for a new prompt.
3. Set the schedule. Use **Manual** if you want to collect an initial answer before enabling recurring monitoring.
4. Add groups or personas if you want to organise the question now; both are optional.
5. Save the prompt.

For example, an outdoor clothing brand could monitor: “Which waterproof jackets are best for commuting by bike?” This is an illustrative question; use a category and audience relevant to your own business.

## Collect and inspect an answer

Choose **Run now** for the prompt. In a new workspace, the walkthrough leads you to this action after the prompt is saved. A queued or running status means collection is in progress. An answer and its analysis can become available separately.

Open **View results** to read the answer, then inspect the brand mentions, sentiment, and evidence. Use the answer history to compare later collections of the same question.

If no answer appears, follow [the missing-answer checks](/docs/troubleshooting/missing-answers). Avoid submitting repeated runs while the first request is still active.

## Build a useful monitoring set

One question gives you one narrow view of AI search. Add questions covering your main categories, customer needs, and competitor comparisons. [Choose prompts deliberately](/docs/getting-started/choose-prompts) and [group them by topic](/docs/workspace/groups-and-personas) so you can interpret the results separately.