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

# Understanding Responses

> How Superlines collects and analyzes AI responses.

Every time Superlines tests a prompt against an AI engine, it captures the full response and runs it through analysis.

## What is a response?

A **response** is the full text answer that an AI engine gives to a tracked prompt. Superlines stores and analyzes each response to extract:

* **Brand mentions** — does the response name your brand or any competitor?
* **Website citations** — does the response reference any URLs?
* **Sentiment** — is the tone positive, neutral, or negative?
* **Content structure** — headers, lists, links, and formatting

## How responses are collected

<Steps>
  <Step title="Prompt submission">
    Your tracked prompts are sent to each enabled AI engine.
  </Step>

  <Step title="Response capture">
    The AI engine generates a response. Superlines captures the full text (text, HTML, and markdown).
  </Step>

  <Step title="Storage">
    Responses are stored in BigQuery for analytics.
  </Step>

  <Step title="Analysis">
    Brand detection and citation extraction run automatically.
  </Step>
</Steps>

<Note>
  Responses are collected daily for active prompts. The testing frequency depends on your plan.
</Note>

## Brand detection

Superlines detects brand mentions using:

* Your **brand name** (normalized for consistent matching)
* **Brand name variations** you've configured
* **Brand exclude list** to avoid false positives
* **Competitor names** and their aliases

For example, if your brand is "Nike Running" and you've added "NRC" and "Nike Run Club" as variations, all three will be detected as mentions of your brand.

## Where to view responses

| What you want              | Where to find it                               |
| -------------------------- | ---------------------------------------------- |
| All responses for a prompt | Click any prompt → Response detail view        |
| Full response text         | Response Analysis page                         |
| Brand mention highlights   | Response detail view (highlighted in response) |
| Citation links             | Citations Analysis page                        |
| Sentiment breakdown        | Brand Sentiment page                           |

## Response metrics per prompt

Each prompt in the **Tracked Prompts** table shows:

| Column                          | Meaning                                |
| ------------------------------- | -------------------------------------- |
| **Total responses**             | How many AI responses have been tested |
| **Brand mentions**              | How many responses mention your brand  |
| **Brand Visibility %**          | Mentions / total responses             |
| **Citations**                   | How many responses cite your website   |
| **Branded Response Rate (BRR)** | % of responses that mention any brand  |

<Tip>
  **BRR** tells you how "brand-focused" a topic is. A high BRR means brands are commonly mentioned for this query — making it competitive but valuable to win.
</Tip>

## Why numbers may fluctuate early on

When you first create a brand, numbers may change rapidly — especially in the first hour as initial responses come in. This is normal:

* **Small sample sizes** make percentages volatile
* **Each new response** has a larger relative impact
* Data **stabilizes** after 24–48 hours as more responses are collected

Give your brand at least 48 hours of data collection before drawing conclusions.