> ## Documentation Index
> Fetch the complete documentation index at: https://docs.get3rd.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Setting up your prompts

> Prompts are the foundation of everything 3RD tracks — learn how to write, generate, and categorize them so the insights that follow are meaningful.

Prompts are the questions 3RD sends to the AI models on your behalf, every day. The closer they are to how real buyers talk to AI, the more your dashboard reflects reality.

3RD suggests prompts based on your topics — and those suggestions are designed to reflect how real people actually use AI assistants, grounded in research covering hundreds of millions of users. You're not starting from a blank page or guessing at phrasing.

<Tip>
  Write prompts **conversationally and with context** — "what's the best project tool for a 5-person design studio?" beats the bare keyword "project tool". People don't type keywords into ChatGPT; they ask questions.
</Tip>

## Where good prompts come from

<AccordionGroup>
  <Accordion title="Rephrase your search queries" icon="magnifying-glass">
    Take the queries you already rank for in Google Search Console and rephrase them as questions. If you [connect GSC as a data source](/integrations/data-sources), 3RD does this for you — it turns queries you get impressions for into conversational prompts worth tracking.
  </Accordion>

  <Accordion title="Accept 3RD's suggestions" icon="wand-magic-sparkles">
    3RD suggests prompts based on your brand, website, and topics. Accept the good ones, discard the rest, and add your own on top.
  </Accordion>

  <Accordion title="Ask your sales and support teams" icon="headset">
    The questions prospects actually ask before buying — "does X integrate with Y?", "is X reliable for enterprise?" — are exactly the questions they now ask AI.
  </Accordion>
</AccordionGroup>

## Branded vs. unbranded prompts

You want both — they measure different things:

| Type          | Example                                | Measures                                   |
| ------------- | -------------------------------------- | ------------------------------------------ |
| **Branded**   | "Is Acme good for enterprise?"         | How AI talks about you when asked directly |
| **Unbranded** | "What's the best CRM for small teams?" | Whether you get discovered at all          |

3RD detects the type automatically by checking whether the prompt names your brand.

## Categorize by commercial value

Assign each prompt a type so prioritization downstream is automatic. 3RD pre-categorizes every prompt, and you can adjust any classification manually:

| Tier             | What it means                                                                                                                                                                                            | Recommended visibility target |
| ---------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------- |
| **Must-Win**     | Direct commercial impact — the questions where the user is closest to a decision, and where AI determines which solutions get highlighted, compared, and prioritized. This is where market share shifts. | 75%                           |
| **Core**         | Shapes preferences and shortlists — the research phase, where users structure their decision landscape and your brand becomes associated with the category.                                              | 50%                           |
| **Foundational** | Builds AI understanding and category clarity — the context AI uses to understand your brand and your role in the market.                                                                                 | 35%                           |

<Note>
  Start with **10–15 prompts you genuinely care about** rather than maxing out your plan on day one. It's easier to read clean signal from a focused set, and you can always add more from the Prompts page.
</Note>

## Related

* [Topics, markets & audiences](/setup/topics-markets-audiences) — the structure your prompts live in
* [Data sources](/integrations/data-sources) — seed prompts from real query data