# Sentiment

Inspect the tone of tracked answers and find the evidence behind positive or negative results.

Canonical: https://docs.equalai.io/analytics/sentiment
Markdown: https://docs.equalai.io/analytics/sentiment.md

Open **Reports → Sentiment** to investigate how captured answers describe your brand. Sentiment is an automated classification of answer tone, not a customer survey or a measure of sales performance.

## Review the overview

1. Set the date range and **All filters**.
2. Start on **Overview** and compare **Sentiment over time**, **Share of sentiment**, and **Sentiment distribution**.
3. Use **Net sentiment by model** to identify model differences.
4. Check **Brand reach** separately: competitor reach shows mentions, not independently scored competitor sentiment.

![EarlySEO Sentiment report header showing its summary mix, report views, and search and date filters](https://docs.equalai.io/images/guides/sentiment.webp)

*EarlySEO example snapshot of the report's top controls. **1.** Summary and sentiment mix. **2.** Switch report views. **3.** Search, date and filters. Use the sections below these controls to investigate the displayed classifications.*

The sentiment categories are **Very Positive**, **Positive**, **Neutral**, **Negative**, **Very Negative**, and **Unknown**. Unknown means no usable classification; do not treat it as evidence of neutral language.

## Understand the net score

Positive categories contribute +1, negative categories −1, and neutral/unknown categories 0. EqualAI divides the total by all responses in scope and multiplies by 100, then rounds. Thus unknown responses reduce the magnitude of the score without adding positive or negative evidence.

A score of +100 means every response in that scope contributed positively. A score of zero can mean neutral/unknown answers or a balance of positive and negative answers. Read the distribution before deciding which explanation applies.

The time chart contains days with responses. A gap does not establish stable sentiment during the missing period.

## Investigate individual answers

- Open **Details** to inspect the prompt, model, answer, classification, and available sentiment explanation.
- Use sentiment, topic, search, and sorting controls to find the answers that need review.
- Open **Topics** to compare question groups.
- Open **By Tag** to compare your own tracking segments.
- Use **Export CSV** for an external review of the report data.

Read the original answer before acting on a negative label. Quoted criticism, comparisons, and mixed-tone answers can require human interpretation.

## Troubleshooting

- **Many answers are Unknown:** inspect response evidence and model coverage instead of assuming neutral perception.
- **A model score is volatile:** compare its response count and period with the other models.
- **Net score and distribution seem inconsistent:** unknown answers remain in the net-score denominator.
- **Competitor sentiment is missing:** Brand reach is a mention comparison, not a competitor tone report.

## Related guides

[Responses](https://docs.equalai.io/analytics/responses) · [Metrics reference](https://docs.equalai.io/analytics/metrics) · [Reputation](https://docs.equalai.io/analytics/reputation)