> ## 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.

# Reading your results

> How to interpret your numbers: when to trust them, mentions versus citations, rates versus shares, and answers with no sources.

Every figure in RankRaft AI comes from stored answers, but a number on its own can still mislead. These are the habits that keep your conclusions sound.

## Give it a week

AI engines give different answers to the same question from one day to the next. Your first scan is your [baseline](https://docs.rankraft.ai/getting-started/glossary.md), not a verdict. Wait for about a week of daily scans before judging a prompt, an engine or a competitor.

* Figures based on very few answers are greyed out as **Low sample · N answers**. Treat them as hints.
* Every Visibility Score carries a **± margin**. Two brands within each other's margins are *tied within margin*, so don't read anything into their order. See [Visibility Score](https://docs.rankraft.ai/analytics/visibility/visibility-score.md#margin).
* Longer ranges, more prompts and more engines all narrow the margin.

## Compare like with like

* **Same length of period.** A 30-day score covers different answers from a 7-day score, and prompt coverage in particular rises with longer ranges.
* **One engine at a time when it matters.** Each engine draws on different sources and names different brands. A pooled score can hide a strong engine and a weak one. Break results down by engine before deciding where to act.
* **Check the chart markers.** **P** marks a day where some answers couldn't be collected. **C** marks a day where you changed your setup, such as your prompts or competitors. A step change on a **C** day is usually the change, not the market.

## Being named and being cited are different signals

An answer can name your brand, link to your site, do both or do neither. Each combination points to a different next step.

| | Linked to your site | Not linked |
| - | - | - |
| **Names you** | The best outcome: AI recommends you and sends readers to you | AI knows you, but relies on other sites as evidence. Work on the pages it cites instead (see [Where to get featured](https://docs.rankraft.ai/analytics/citations/citations-and-sources.md#where-to-get-featured)) |
| **Doesn't name you** | AI uses your content as a source but recommends someone else. Make your pages say clearly who you are and what you offer | You're absent. Start with the prompt's competitors and the sources behind their answers |

Mention rate and citation rate are measured separately for this reason, and the [Visibility Score](https://docs.rankraft.ai/analytics/visibility/visibility-score.md) weights them 45% and 35%.

## Rates and shares answer different questions

* **Rates** (mention rate, citation rate) are per answer: *in what share of answers did this brand appear?* Every brand is measured against the same answers, so rates don't add up to 100% across brands. Five brands can each be named in 60% of answers.
* **Shares** (share of voice, citation share) split a total between brands or sites: *of all the mentions, or all the links, how many went to this one?* Across every brand on your list, shares add up to 100%. The leaderboard only shows confirmed competitors, so the shares you see can add up to less.

A brand can gain mention rate and lose share of voice in the same week, if rivals gained faster.

## Answers with no sources

Engines decide for themselves whether to search the web before answering, as they do for a real user. When an engine answers from what it already knows, the answer cites nothing. That is normal behaviour, not a sign your site was ignored.

* Questions that ask for a recommendation or a comparison are searched more often than questions that ask for a definition.
* A prompt whose answers rarely cite anything gives citation rate little to work with. Mentions are the signal to watch there.

## When Google shows no AI answer

About 30% of Google result pages show no AI Overview at all. Those runs appear as **No AI answer shown** and are left out of every rate, because they say something about the question, not about you. See [How we measure](https://docs.rankraft.ai/getting-started/how-we-measure.md#when-an-engine-shows-no-answer).

## Average spot only covers answers that name you

**Average spot** is where you appear in answers that name you: 1 means first. It says nothing about answers that leave you out, so always read it next to mention rate. A brand named once, first, has a perfect spot and very little visibility.

## Where to go next

> [card] [Visibility Score & metrics](https://docs.rankraft.ai/analytics/visibility/visibility-score.md)
>
> The exact formulas, with a worked example.

> [card] [Diagnose with logs and answers](https://docs.rankraft.ai/agent-analytics/diagnose-with-logs.md)
>
> Find out why a page is read by AI crawlers but never cited.


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