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

# Sentiment

> A 0-100 score for the tone AI answers use when they name your brand.

Being named is only half the story. An answer can name you as the market leader or as the option with poor support. Sentiment summarizes the tone of the language around your brand's mentions.

## How each mention is classified

For every answer that names a brand, Echod looks at the wording surrounding that brand's mention and classifies it as **positive**, **neutral** or **negative**. Praise ("reliable", "easy to set up", "best for") pushes it positive; criticism ("expensive", "limited", "complaints") pushes it negative; plain description stays neutral.

## How the score is calculated

```text theme={null}
Sentiment = ((positive - negative) / all classified mentions + 1) / 2 x 100
```

The result is rounded to a whole number between 0 and 100:

| Score | Meaning |
| - | - |
| 100 | Every mention is positive |
| 50 | Neutral: positives and negatives balance out, or everything is neutral |
| 0 | Every mention is negative |

Neutral mentions pull the score toward 50, so a brand described mostly in neutral terms will sit close to the middle even if it has a few strong positives.

### Example

Your brand was named in 40 answers: 22 positive, 14 neutral and 4 negative.

```text theme={null}
((22 - 4) / 40 + 1) / 2 x 100 = 72.5, rounded to 73
```

When a brand has no mentions in the period, there is no Sentiment score to show.

## Where you see it

* **Dashboard**: the **Sentiment** row in the Score breakdown.
* **Topic and prompt detail pages**: a gauge per brand with 50 marked as neutral, plus the number of mentions behind it.
* **Chat detail**: the tone of your brand's mention in that single answer, for example "Positive tone 81/100".

![Sentiment gauges for tracked brands on a topic detail page, with Neutral marked at 50](https://mintcdn.com/sherpas-design/qmEl5MUhXLHMeD6J/images/metrics/sentiment-gauge.png?fit=max\&auto=format\&n=qmEl5MUhXLHMeD6J\&q=85\&s=798d25d1283316742b600431079b84be)

## Put it to work

* Sentiment is most useful in comparison. A score of 60 is good if competitors sit at 52, and weak if they sit at 75.
* When Sentiment drops, filter the [chats](https://docs.echod.ai/analyze/chats.md) on a topic or prompt and read the answers that name you. The wording usually points to a specific theme, such as pricing or support, and the cited sources show where that framing comes from.
* Low sample sizes swing easily. A score built on five mentions can move a lot from one week to the next; check the mention count on the gauge before reacting.


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