> ## Documentation Index
> Fetch the complete documentation index at: https://docs.tryxlr8.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# AI Insights Agent

> Ask natural language questions about your experiment data and get instant, AI-powered answers about your AEO performance.

The AI Insights Agent is a conversational assistant built directly into your experiment results. Instead of manually scanning tables and charts, you can ask plain-language questions and receive data-driven answers grounded in your specific experiment and execution results.

The agent is context-aware: it knows which experiment and execution you are viewing, so your questions do not need to include IDs or technical identifiers.

## Accessing the Insights Agent

Open any experiment from the **Experiments** dashboard, then navigate to its results view. The Insights Agent chat panel appears alongside your results data. You can open a conversation at any time while reviewing an execution.

<Note>
  The agent operates within the scope of the experiment and execution you are currently viewing. It does not search the general web or access data outside your XLR8.ai account.
</Note>

## Asking questions

Type your question in the chat input and press **Enter** or click **Send**. The agent responds with an analysis grounded in your experiment data. You can ask follow-up questions to dig deeper — the agent maintains context across the conversation.

### Example questions

The following questions reflect common AEO analysis workflows. Use them as a starting point or inspiration for your own queries.

<AccordionGroup>
  <Accordion title="Model coverage and mention rates">
    * "Which model mentions us most frequently?"
    * "Which models never cite us in this execution?"
    * "How does our mention rate compare across GPT-4o, Perplexity, and Gemini?"
    * "Which query categories have the lowest mention rate?"
  </Accordion>

  <Accordion title="Visibility trends over time">
    * "Compare our visibility across the last 3 executions."
    * "Has our mention rate improved since we updated our content?"
    * "Show me which queries saw the biggest change between execution 1 and execution 3."
  </Accordion>

  <Accordion title="Competitive analysis">
    * "How does our visibility compare to Competitor X in this execution?"
    * "Which models favor our competitors over us?"
    * "Where are we beating Competitor X, and where are we losing?"
  </Accordion>

  <Accordion title="Citation and gap analysis">
    * "What are the main reasons we're not being cited?"
    * "Which queries consistently fail to mention us?"
    * "Are there query categories where we have zero visibility?"
  </Accordion>

  <Accordion title="Sentiment and tone">
    * "What is the overall sentiment toward our brand in this execution?"
    * "Which models describe us most positively?"
    * "Are there any negative themes in how we are being mentioned?"
  </Accordion>
</AccordionGroup>

## Conversation flow

The agent maintains the full history of your conversation within the current session. You can build on previous answers to drill down into specific findings.

<Steps>
  <Step title="Start with a broad question">
    Open with a high-level question, such as "Summarize our visibility results for this execution."
  </Step>

  <Step title="Follow up on specifics">
    Ask follow-up questions to investigate particular models, categories, or competitors: "Why is our mention rate low on Perplexity specifically?"
  </Step>

  <Step title="Request actionable takeaways">
    Close the conversation with an action-oriented question: "What should we focus on to improve our visibility in the next execution?"
  </Step>
</Steps>

## Tips for useful answers

<Tip>
  Be specific about what you want to compare. Questions like "Compare our visibility on GPT-4o vs. Gemini for the Healthcare category" produce more focused answers than "How are we doing?"
</Tip>

<Tip>
  If you are viewing a specific execution, the agent focuses on that execution's data. Switch to a different execution in the results view before asking execution-specific questions.
</Tip>

<Tip>
  Ask the agent to summarize before you dig into details. A summary response often reveals the most important finding and guides your next question.
</Tip>

## Limitations

<Warning>
  The Insights Agent answers questions based on your experiment data only. It cannot browse the web, fetch live search results, or access data from outside your XLR8.ai account.
</Warning>

* Conversations are session-scoped. Starting a new session clears the conversation history.
* The agent focuses on the experiment and execution you are currently viewing. If you need to compare across multiple experiments, navigate to each one separately.
* Answers reflect the data captured at the time the execution ran. The agent does not have access to real-time AI model outputs.

## Related

<CardGroup cols={2}>
  <Card title="Experiments overview" icon="flask" href="/experiments/overview">
    Learn how experiments are structured and how executions are run.
  </Card>

  <Card title="Visibility experiments" icon="eye" href="/experiments/visibility">
    Understand how visibility data is collected across AI models.
  </Card>

  <Card title="Sentiment experiments" icon="face-smile" href="/experiments/sentiment">
    See how sentiment is measured and scored per execution.
  </Card>

  <Card title="Action Center" icon="list-check" href="/intelligence/action-center">
    Turn insights into prioritized recommendations.
  </Card>
</CardGroup>