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

# Win a prompt you are losing

> Take one prompt where competitors are recommended and you are not, and work it until engines mention you.

Pick one prompt from [Content gaps](/docs/geo/content-gaps) with a high opportunity score and work it end to end before you start the next one.

<Steps>
  <Step title="Read the answers">
    Open the prompt on the [Prompts](/docs/geo/prompts) page and, for each engine, note which brands it recommends, in what order and why. The reasons engines give are your content brief.
  </Step>

  <Step title="Read the citations">
    Grounded engines save the pages they cited, which are the sources each engine built its answer from. For more on how they are counted, see [Citations](/docs/geo/metrics/citations).
  </Step>

  <Step title="Get on the shelf">
    Third-party pages in those citations show up in [Shelf space](/docs/geo/shelf-space), so open a ticket for each page that lists competitors but not you.
  </Step>

  <Step title="Publish your own answer">
    Use the [GEO writer](/docs/geo/writer) to draft a page that answers the prompt directly, in the format engines already favor for it, whether that is a comparison, a listicle or a guide.

    Add specific statistics, quotes from named people and links to sources. In the [GEO study](https://arxiv.org/abs/2311.09735), quotes raised visibility in AI answers by 42.6%, statistics by 32.8% and cited sources by 27.7%, while keyword stuffing lowered it.
  </Step>

  <Step title="Make it readable for agents">
    Run [Agent readiness](/docs/geo/agent-readiness) and check the [AI traffic](/docs/ai-traffic/overview) page to confirm crawlers and assistants fetch the new page.
  </Step>

  <Step title="Measure">
    Watch **What changed** over the next scans for **Gained mention** and **Citation gained** on this prompt.
  </Step>
</Steps>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.