> ## 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.

# Understanding conversations

> Conversations is where every metric grounds out in raw evidence — the actual prompts sent, the answers each model returned, and what got cited.

The **Conversations** explorer lets you browse the actual prompts and the answers each AI model gave. Every score in 3RD — visibility, sentiment, position — is an aggregate of these individual answers. When a number moves and you want to know *why*, this is where you look.

## What you can see per answer

For any AI response, Conversations shows:

| Element            | What it tells you                                              |
| ------------------ | -------------------------------------------------------------- |
| **Mention status** | Whether your brand appeared in the answer                      |
| **Sentiment**      | How you were described, with the driving language identifiable |
| **Position**       | Where you landed in the answer's ranked list                   |
| **Competitors**    | Which other brands appeared alongside you                      |
| **Citations**      | Every source the model used to build the answer                |

You can switch between models to compare how ChatGPT, Gemini, Perplexity, and others answered the *same* prompt — often very differently.

## How to use it

<AccordionGroup>
  <Accordion title="When a metric moves" icon="chart-line-up">
    Filter Conversations to the prompt/model/market where the change happened and read the answers. Did a competitor start appearing? Did a new source get cited? Did the framing change? The aggregate tells you *that* something changed; the conversations tell you *what*.
  </Accordion>

  <Accordion title="When sentiment drops" icon="face-frown">
    Open the negative-sentiment answers and find the recurring theme — an outdated pricing claim, a support complaint from a review site, a missing feature. That theme is your fix.
  </Accordion>

  <Accordion title="When you want to understand a competitor's win" icon="user-magnifying-glass">
    Read the answers where they're recommended and you're not. Check which sources the model cited — those sources are your outreach targets.
  </Accordion>
</AccordionGroup>

<Tip>
  Make reading a handful of conversations part of your weekly routine. The metrics compress reality; the answers *are* reality.
</Tip>

## Related

* [Understanding citations](/results/citations) — the sources behind the answers
* [Sentiment](/metrics/sentiment) — how tone becomes a score