Sentiment Analysis: how positively AI describes your brand

Sentiment Analysis classifies how positively or negatively AI models discuss your brand. The analyzer uses the main brand name and the matching names configured for your project or competitor. The page shows classification coverage so you can compare detected mentions with completed sentiment results. Available on Growth plans and above.

With Sentiment Analysis, you can:

• See whether ChatGPT, Perplexity, and Gemini describe your brand positively, negatively, or neutrally
• Track your Net Sentiment Score against competitors over time
• Find the topics and prompts where AI sentiment turns negative, so you know what to address
• Measure how sentiment shifts after a launch, a campaign, or a reputation fix

The Sentiment dashboard brings these signals together in one place: the sentiment distribution, how share of sentiment shifts over time, an at-a-glance donut, and a Net Sentiment Score ranking that puts you next to your competitors.

The Overview tab below is where you read all of this at once.

What it does

• Classifies detected mentions as very positive, positive, neutral, negative, or very negative
• Shows how many detected mentions have a sentiment classification
• Calculates a Net Sentiment Score ranging from -100% (all negative) to +100% (all positive)
• Identifies sentiment topics: the themes AI discusses about your brand (shown as a treemap)
• Compares sentiment across your brand and all competitors
• Offers 5 analysis tabs: Overview (charts + rankings), Details (individual records), Topics (treemap), By Tag (per-tag breakdown), Pages (sentiment of the content of cited pages, showing negatives by default)
• Every tab's table can be exported to CSV or XLSX with the Export button, respecting the active filters
• Creates a branded PDF report with the current filters, including the sentiment snapshot, trend, competitor benchmark, classification coverage, topics, tags and detailed findings

How to use it

1. Navigate to Sentiment in the sidebar (under Reports)
2. Select a brand from the dropdown: "All Brands" for aggregated view, your brand, or a specific competitor
3. Use the Overview tab for charts showing sentiment trends over time
4. Switch to Topics to see a treemap of what AI says about the selected brand
5. Use By Tag to compare sentiment across different prompt groups
6. Use Details for individual sentiment records with comments and topic labels
7. Select Export PDF in the page header to create a shareable report. LLM Pulse generates it in the background and adds it to Exports in the user menu.

Add context that is specific to your brand

Words can sound positive or neutral while still being harmful to a particular brand. For example, a person may not want AI responses to frame them as a former politician. You can add a rule that treats that framing as negative, even when the response uses neutral language.

Open Project Settings > Sentiment Rules, or use the Sentiment rules button at the top of the Sentiment page. You can save up to 5 plain-text rules, with a maximum of 200 characters each. Keep each rule focused on one situation and state the sentiment you expect.

These rules override the general scoring guidance only when they apply to your own brand. They do not change mention detection or sentiment for competitors. A saved rule applies to new and re-run analysis; it does not rewrite existing results automatically.

Tips & notes

• Available on Growth plans and above
• Weekly negative sentiment email: turn it on in your project settings to get an alert covering both the negative AI responses of the last 7 days and the cited pages whose content criticizes your brand. A page is included when an AI model cited it during that week, so the list reflects what the models are pointing at right now.
• Sentiment processing can take up to 24 hours after prompt execution (uses OpenAI Batch API)
• Matching names are used for sentiment analysis as well as mention detection
• Custom sentiment rules can adjust how new and re-run analysis interprets business context for your own brand
• Sentiment percentages and Net Sentiment Score use classified mentions only
• A detected mention can be absent from sentiment totals while analysis is processing or if no classification is available
• Net Sentiment Score = (positive + verypositive - negative - verynegative) / classified mentions
• Each sentiment record includes a comment explaining the AI's reasoning
• Sentiment analysis covers your brand AND all tracked competitors