> For the complete documentation index, see [llms.txt](https://docs.airops.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.airops.com/insights/prompts/fact-check.md).

# Fact Check

Check whether AI answers describe your brand accurately using your own pass and fail criteria.

Fact Check evaluates tracked AI answers against facts and criteria you define. Use it to catch incorrect pricing, product claims, or descriptions of the customers your brand serves.

## How to configure a fact check

1. Open **Insights → Prompts → Fact Check** and select your Brand Kit.

   <figure><img src="https://3762890407-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FX2n5yPRPynbnWuO4SH0M%2Fuploads%2Fgit-blob-499838505330825f9b02b3069fa33d60a6971b04%2Ffact-check-01-list.png?alt=media" alt="Fact Checks tab on the Prompts page listing four evaluations for the SpaceOps Brand Kit with their assigned prompts, related topics, answers evaluated, and answer pass rates, below the Date, Region, and Platforms filters"><figcaption><p>The Fact Checks tab lists each check with its assignment and answer pass rate for the selected date range.</p></figcaption></figure>
2. Choose **Create with Quill** to draft checks from your brand context, knowledge base, website, and tracked prompts, or choose **Create manually**.
3. Review the name, supporting context, and **Pass** and **Fail** criteria. Check the source for each fact before saving.
4. Choose the prompts or topics the check should evaluate.

   <figure><img src="https://3762890407-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FX2n5yPRPynbnWuO4SH0M%2Fuploads%2Fgit-blob-26baec37fd3268ec276c310c978db43086c61c8a%2Ffact-check-02-create.png?alt=media" alt="New evaluation form for a Starting price check assigned to the Pricing &#x26; Plans topic, with Pass and Fail conditions about the $1,200 per satellite starting price and an empty Test samples panel"><figcaption><p>Name the check, choose the prompts or topics it covers, and write its Pass and Fail conditions.</p></figcaption></figure>
5. Review the draft and save the evaluation.

Creating and editing checks requires evaluation-create permission. **Create with Quill** also requires permission to use Quill. Quill opens a new chat and sends a starter request; review its proposed checks before creating them.

## Read the results

Each evaluated answer receives a **Pass**, **Fail**, or **Irrelevant** result. You write the pass and fail criteria; Irrelevant is the fallback for answers the criteria do not apply to.

Use the date range and filters to compare results over time, then open an evaluation to inspect the answers behind its results. Pass and fail rates use relevant answers, excluding irrelevant results. For example, 60 passes, 20 failures, and 20 irrelevant answers produce a 75% pass rate across 80 relevant answers.

<figure><img src="https://3762890407-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FX2n5yPRPynbnWuO4SH0M%2Fuploads%2Fgit-blob-5c08128939871ceeb7c1726871ee05e9d26e8be8%2Ffact-check-03-results.png?alt=media" alt="Starting price evaluation opened on the prompt how much does SpaceOps cost, listing evaluated answers by date with each platform&#x27;s Pass, Fail, or Irrelevant result and reasoning"><figcaption><p>Open a prompt in an evaluation to see each answer's result and the reasoning behind it.</p></figcaption></figure>

## Edit an existing check

Open the evaluation's actions menu and choose **Edit**. Update its name, context, criteria, or assignment, then save.

Editing criteria does not re-evaluate answers already judged. Historical results keep their original verdicts, so a period spanning an edit can include results from different criteria. Changing the assigned prompts or topics also changes the scope of future evaluations.

## Task usage

Recurring evaluations use **1 task per 10 answers evaluated**. Review consumption on the Fact Check and Usage screens. Evaluation usage is separate from the tasks used to collect the answers.

## Example

Create a pricing check using your current pricing page as context. Define a pass as an answer quoting the current starting price and a fail as an answer quoting a different starting price. Answers that do not discuss pricing can be irrelevant. Review failed answers to see which claims need attention.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.airops.com/insights/prompts/fact-check.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
