> For the complete documentation index, see [llms.txt](https://passion-io.gitbook.io/help/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://passion-io.gitbook.io/help/make-your-ai-answer-well/check-your-ai-is-ready.md).

# Check Your AI Is Ready

Run your AI against real questions and see where they answer from your material, where they're weak, and whether your boundaries hold.

> **Evaluation** is available on certain plans and app types. If you don't see it in the AI block in your sidebar, please check our pricing page for more information.

## In This Article

* [Run Your First Check](#run-your-first-check)
* [Read the Results](#read-the-results)
* [Add Your Own Questions](#add-your-own-questions)
* [Compare Against Earlier Checks](#compare-against-earlier-checks)

## Run Your First Check

1. In the sidebar, open the AI block and select **Evaluation**.
2. Start a check. Before you've run one, the page reads "You haven't checked your AI yet".
3. Wait while your AI answers each question. **Latest check** then shows how they answered.

Run a check before you share your AI with anyone, and again after any substantial change to your **Resources** or **Behavior**.

## Read the Results

Each question gets a verdict:

* **Working well** — a good answer, grounded in what you gave them.
* **Weak answer** — they answered, but not well enough to stand in front of a client.
* **Needs attention** — the answer is a problem. Start here.
* **Not scored** or **Couldn't check** — no verdict for that question this time.

Each result also says where the answer came from — **From your material** or **Not from your material**. An answer marked "Not from your material" is your AI filling a gap you haven't covered, which is the signal to add a resource.

Boundary questions report separately as **Boundary held** or **Boundary missed**. A missed boundary is the most urgent thing on the page: use **Review boundaries** to tighten the rule that let it through.

Select **Show answer** on any row to read the full response, and **Hide answer** to collapse it again.

## Add Your Own Questions

The stock questions won't cover what your clients actually ask. Under **Questions you add**:

1. Add the questions you get most, in the words clients use.
2. Include the awkward ones — pricing, refunds, and anything you'd rather your AI declined.
3. Remove a question with **Remove question** when it stops being relevant.

Your questions run alongside the standard ones on every later check, so the same weak spots get retested.

## Compare Against Earlier Checks

**Past checks** keeps your history. Open an earlier check to compare it against the latest one and confirm a fix actually worked. If a question moved from **Weak answer** to **Working well** after you added a resource, that's your evidence the resource was the missing piece.

## Related

* [Add Resources to Your Library](/help/make-your-ai-answer-well/add-resources.md)
* [Set Boundaries for Your AI](/help/make-your-ai-answer-well/set-boundaries.md)
* [How Your AI Decides What to Say](/help/make-your-ai-answer-well/how-your-ai-decides.md)

For more assistance, please contact us.