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

# How metrics are calculated

> The exact formulas behind the Visibility Score, mention rate, citation rate, coverage, position, share of voice and margins.

Every number in RankRaft AI is calculated from the stored answers in the selected date range. Nothing is modelled or estimated. Every rate counts only answers that were actually shown: when Google shows no AI Overview for a question, that run is left out rather than counted as "not mentioned".

## Visibility Score

A single number from 0 to 100:

$$
\text{Score} = 100 \times (0.45\,M + 0.35\,C + 0.20\,P)
$$

where *M* is your mention rate, *C* your citation rate and *P* your prompt coverage, each between 0 and 1. The result is rounded to the nearest whole number.

| Component | Weight | Why |
| - | - | - |
| Mention rate | 45% | Being named is the outcome that matters most. |
| Citation rate | 35% | A link is evidence AI trusts your site, and it's what sends traffic. |
| Prompt coverage | 20% | Rewards showing up across many questions rather than dominating one. |

Position isn't part of the score on purpose. It's only defined for answers that list brands in order, so including it would penalise brands mentioned in prose.

## Component metrics

| Metric | Formula |
| - | - |
| **Mention rate** | answers that name you (or an alias) ÷ all answers |
| **Citation rate** | answers that link to at least one page on your domain ÷ all answers |
| **Prompt coverage** | prompts where you were named at least once ÷ prompts asked in the range |
| **Average spot** | in each answer, the brands found are numbered by first appearance (1 = named first). Averaged over answers that mention you, to one decimal place |
| **Share of voice** | your mentions ÷ all mentions of you and your confirmed competitors |

### Example

In a week with 10 prompts on 3 engines (210 answers): you're named in 63 answers, cited in 21, and named at least once for 6 of the 10 prompts.

* Mention rate = 63 ÷ 210 = 30%
* Citation rate = 21 ÷ 210 = 10%
* Prompt coverage = 6 ÷ 10 = 60%
* Visibility Score = 100 × (0.45 × 0.30 + 0.35 × 0.10 + 0.20 × 0.60) = **29**

## Margin

Every score carries a 95% interval, shown as **± N** and as a shaded band on trend charts:

$$
\text{margin} = 100 \times 1.96 \times \sqrt{\frac{p(1-p)}{n}}
$$

where *p* is your mention rate (held between 5% and 95%) and *n* is the number of answers. The margin is never shown as less than 1 point. More prompts, more engines and longer date ranges all narrow it.

**Tied within margin:** on the leaderboard, two neighbouring brands whose scores differ by no more than their combined margins are marked as tied. Don't read anything into the order of tied brands.

## Competitors

Competitors are scored with exactly the same formula, over the same answers. A competitor is *cited* when an answer links to its domain or a subdomain of it. **Rank** orders you and your confirmed competitors by Visibility Score.

## Importance

Prompts and gap prompts are tagged by importance, based on your mention rate for that prompt:

| Importance | Your mention rate |
| - | - |
| High | 0% |
| Medium | Below 40% |
| Low | 40% or more |

## Small samples

Figures based on very few answers are greyed out and labelled **Low sample · N answers**. Treat them as a hint, not a result.

## Why past numbers can change

The scores for a date range are recalculated from the stored answers whenever you look. They change if you:

* **Confirm or archive a competitor.** The leaderboard, rank and share of voice include confirmed competitors only.
* **Add an alias.** Earlier answers that used that name now count as mentions.
* **Change the range.** A 30-day score covers different answers from a 7-day score.


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