> ## 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.

# Analysis Process

> How Superlines collects, processes, and analyzes your AI search data.

Understanding how data flows through Superlines helps you interpret your metrics and make better decisions.

## Daily analysis cycle

Superlines runs a daily analysis cycle for each active brand:

```
Tracked Prompts → AI Engine Testing → Response Collection → Analysis → Metrics & Insights
```

<Steps>
  <Step title="Prompt selection">
    Active prompts for the brand are queued for testing.
  </Step>

  <Step title="AI engine testing">
    Each prompt is sent to all enabled AI engines.
  </Step>

  <Step title="Response collection">
    Full responses are captured (text, HTML, citations).
  </Step>

  <Step title="Brand detection">
    Responses are scanned for brand mentions using the brand name, variations, and competitor names.
  </Step>

  <Step title="Citation extraction">
    URLs referenced in responses are parsed and tracked.
  </Step>

  <Step title="Fan-out tracking">
    Web searches performed by AI engines are captured.
  </Step>

  <Step title="Metric calculation">
    Brand Visibility, SOV, Citation Rate, and other metrics are computed.
  </Step>

  <Step title="Insight generation">
    Trends, alerts, and recommendations are created.
  </Step>
</Steps>

<Note>
  This process runs **daily** for active prompts. New prompts are queued and tested within 24 hours of being added.
</Note>

## When data becomes available

| Event                 | Timeline                             |
| --------------------- | ------------------------------------ |
| Brand created         | First results within 1–2 hours       |
| Prompt added          | Tested within 24 hours               |
| Prompt paused/resumed | Takes effect immediately             |
| Engine added/removed  | Takes effect from next testing cycle |
| Data stabilization    | 48 hours for reliable metrics        |

## Why initial numbers may fluctuate

When you first create a brand or add new prompts, you may see numbers change rapidly:

* **First hour** — initial results come in, percentages swing with small sample sizes
* **First 24 hours** — more responses are collected, metrics begin to stabilize
* **48+ hours** — enough data for reliable metrics and trend detection

This is expected behavior. Superlines needs a representative sample of AI responses before metrics accurately reflect your brand's visibility.

## How metrics are aggregated

| Level           | What it shows                                            |
| --------------- | -------------------------------------------------------- |
| **Per prompt**  | How you perform on a specific question                   |
| **Per engine**  | How you perform on a specific AI platform                |
| **Per brand**   | Overall brand performance across all prompts and engines |
| **Per label**   | Performance for a group of prompts sharing a label       |
| **Cross-brand** | Comparison across all your brands                        |

## Trend detection

Superlines automatically detects and surfaces:

* **Week-over-week changes** in core metrics
* **Prompt-level alerts** when visibility drops significantly
* **Competitor movements** — new competitors gaining traction
* **Content opportunities** — topics with high volume but low visibility

These are surfaced as **Signals** on prompts (Trophy, Star, Alert, etc.) and as **Next Actions** on the dashboard.

## Data retention

* All response data is retained for the lifetime of your subscription
* Historical trends are available from the moment you start tracking
* Paused prompts retain their historical data
* Deleted prompts retain aggregated metrics but remove raw response data