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

# Sentiment

> Read how AI engines describe your brand: the 0 to 100 sentiment score, what they praise and criticize, how you compare with one rival, and which attributes you win or lose on.

> [!NOTE]
>
> **Plans:** All plans · **In the app:** Sidebar → Analytics → Sentiment

The Brand Sentiment page answers: **how do AI engines talk about my brand, on what, and is it getting better?** Every number comes from passages quoted from real answers, so you can always open the sentences behind a score.

In the sidebar, open **Sentiment** under **Analytics**. For how passages are extracted and labeled, see [Sentiment metrics](https://docs.usequotable.com/metrics/sentiment.md).

## How the score works

Each passage about a brand is positive, neutral or negative. The sentiment score turns those counts into a number from 0 to 100:

```text theme={"system"}
sentiment score = (100 × positive + 50 × neutral) / all passages
```

50 means balanced or purely factual coverage. The score ring is green from 60, amber from 40 to 59 and red below 40. Next to the score, a signed change compares it with the period of the same length right before your date range. Nothing shows when either period has no passages or the score did not move.

## Controls

The toolbar sits above the cards.

| Control | What it does |
| - | - |
| **Date range** | Today, Last 7 days and other presets, All time, or custom. Default is the last 30 days. |
| **Filters › Models** | One engine or several. |
| **Filters › Locations** | Restricts the answers to the locations you pick. |
| **Filters › Prompt type** | **Branded prompts** or **Unbranded prompts**. Decided by the prompt text, never the answer. |
| **Filters › Competitors** | Chooses the rivals shown in comparisons. Your brand's own figures do not change. |
| **Export** | Downloads a CSV of positive, neutral and negative passage counts per brand and aspect, for the range and models in view. |

The filters stay in the page URL, so a shared link opens the same view.

## The page, top to bottom

### Brand perception

"How AI answers describe" your brand, across the full width.

* **Sentiment score** ring with its change, and the split of your passages: **% pos**, **% neu**, **% neg**.
* A summary sentence: "AI answers credit *Brand* for …" and "The criticism centres on …", naming up to four aspects each. Each aspect is a link that opens the answers behind it, filtered to praise or criticism.
* The footer lists a score per engine, highest first, for every engine with passages in the date range. It ignores the **Models** filter, so you always see where engines disagree.

### Sentiment score

"Split and net score by day", or **by week** when the range is longer than 14 days. Each bar is your positive, neutral and negative share for that period, and the line is the net score.

Turn on **Compare competitors** to swap the bars for one net-score line per brand: yours and each tracked competitor with passages in the range. Hover a name in the legend to bring its line forward.

### Head-to-head

"Visibility and what each brand is known for", against one competitor. Pick the rival from the menu in the card header. It lists your tracked competitors, narrowed by the **Competitors** filter when it is set.

| Part | What it shows |
| - | - |
| Two percentages and a bar | Visibility: the share of answers that mention each brand |
| **Known for** | Up to four aspects with more praise than criticism, and up to two with more criticism than praise, for each side |
| Footer | **Favorability** (each brand's sentiment score), **share of voice**, and "Mentioned more on *N* of *M* models" |

With no competitor tracked, the card asks you to add one.

### Attributes

"What AI weighs when it compares" your brand. A treemap with one tile per aspect. Use the switch in the header for the **Attribute map** or the **Ranked list**.

* **Size** is the aspect's share of every passage in view: positive, neutral and negative, across every brand the answers discuss. When the **Competitors** filter is set, only your brand and the chosen rivals count.
* **Color** is your brand's own standing on that aspect. A rival's result never changes it.

| Standing | Rule, over your brand's praise and criticism |
| - | - |
| **Praised** | Praise, with less than a tenth criticism |
| **Mixed** | At least a tenth criticism. Also any criticism with fewer than 4 passages. |
| **Neutral** | Your brand is mentioned here, but only neutrally |
| **Criticized** | At least a third criticism, out of at least 4 passages |
| **Not in the conversation** | Answers discuss the aspect, but never about your brand |

Neutral passages are left out of the standing rules. The legend gives each standing's share of the conversation, adding up to 100%, and the number of attributes. Hover a tile for its counts, for example "Ranked 2 of 9 brands praised on it." The **Ranked list** shows the same tiles with **Attribute**, **Top rival** (the brand praised most on it), **Share** and **Standing**.

### Strengths and weaknesses

"The aspects AI praises and criticises most about your brand." Switch between **Strengths** and **Weaknesses**; each tab shows its count and up to six aspects, ranked by mentions.

* Hover a row to read a quote from an answer.
* Click a row to open those answers.
* On a weakness, **Respond** opens a brief to "Counter" it, with the quote as evidence and a suggested title you can edit. **Write with the agent** opens [Quotable Agent](https://docs.usequotable.com/actions/quotable-agent.md) with the brief in the message box. Nothing is sent until you send it.

## Reading the answers

Clicking an aspect anywhere on the page opens the matching answers in a viewer. Each answer appears in full, with its positive, neutral and negative passages outlined. The viewer shows the engine, the brands named and the cited sources. Use the arrows to move between answers. The list uses the same dates, models, locations and prompt type as the number you clicked.

## FactCheck

If your workspace shows the **FactCheck** tab next to **Sentiment**, it checks what AI says about your brand against facts you approve. The date range and models carry over when you switch tabs. See [FactCheck](https://docs.usequotable.com/analyze/factcheck.md).

## What to do with it

| You see | What it means | What to do |
| - | - | - |
| A weakness in **Brand perception** | Engines repeat the same criticism across answers | Open it and read the quotes, then use **Respond** on that weakness to brief a page that answers it with evidence |
| A large **Criticized** tile | A heavily discussed aspect where criticism is the story | Fix it first. Its size means it shapes most answers. |
| A large **Not in the conversation** tile | The category discusses it, and you are absent | Publish a page about that aspect with your brand named. Presence comes before praise. |
| A rival **Known for** an aspect you are not | They own that part of the comparison | Build comparison content on that aspect |
| One engine far below the others | That engine reads different sources | Check its citations on [Citations](https://docs.usequotable.com/analyze/citations.md) |

After you publish, keep the same prompts tracked and watch the **Sentiment score** card. Set **Prompt type** to **Unbranded prompts** to see how engines describe you when the question does not name you.

## Next steps

> [card] [Sentiment metrics](https://docs.usequotable.com/metrics/sentiment.md)
>
> Polarity rules, aspects and the score.

> [card] [Visibility by topic](https://docs.usequotable.com/analyze/overview.md#visibility-by-topic)
>
> Which topics each competitor owns.

> [card] [Quotable Agent](https://docs.usequotable.com/actions/quotable-agent.md)
>
> Turn a weakness into a page.


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