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

# How does AI brand sentiment analysis work in chatobserver?

> Interpret positive, neutral, and negative brand sentiment, inspect highlighted passages, and compare competitor descriptions.

Sentiment analysis describes the tone of statements about a brand in collected AI answers. Use **Sentiment** to review patterns, then open individual answers to read the passages behind a positive, neutral, or negative classification.

## Read sentiment in context

A favourable description can include a limitation. For example, “a strong option for larger teams, but expensive for a solo consultant” contains both a strength and a trade-off. The audience and use case matter when interpreting that language.

Sentiment is not the same as visibility. A brand can appear frequently and be described critically, or appear rarely but receive strong recommendations. Review both measures together.

## Inspect highlighted passages

Open an answer from **Prompts**. Your brand's positive and negative passages use distinct highlight colours. Turn on **Highlight other brands** to inspect competitor sentiment as well. Read beyond the highlighted phrase to include conditions, comparisons, and negations in the sentence.

A name-only highlight identifies a mention; it does not prove positive or negative sentiment. An answer with no highlighted passage may have no matching sentiment evidence, or its analysis may still be pending.

## Find useful patterns

Compare recurring descriptions across related questions and dates. Look for consistent associations with features, pricing, audience, geography, or limitations. Use these findings to identify positioning questions worth investigating, rather than treating a single classification as a verdict on your brand.

For unexpected results, follow [sentiment troubleshooting](/docs/troubleshooting/sentiment-highlights). To investigate the material behind an answer, review [source citations](/docs/analysis/source-citations).