TL;DR — Your overall sentiment score may look fine while a specific product area or topic is being destroyed in AI responses. Go to Insights > Sentiment Analysis and apply the Topic or Tag filter to isolate each area of your brand — any topic scoring 15+ points below your average needs investigation. Use the GEO Matrix to identify the exact prompts driving the negative signal, then read those responses in Visibility > Prompts to see the specific language causing the damage. Pro tip: tag your prompts by product line or strategic theme so topic-level sentiment analysis becomes an instant report card for each area of your business.
The Question
“Is AI generating negative sentiment about my brand on specific topics?”Broad sentiment averages can obscure critical issues. Your overall AI sentiment score may be healthy at 65 out of 100, but underneath that average, a specific product line might be scoring 28, or a particular use case might be triggering consistently negative AI-generated descriptions. Users asking about your brand in those specific contexts are receiving a damaged narrative — and you cannot see it from the aggregate score alone. Topic-level sentiment analysis lets you move from “how does AI feel about us overall?” to “where specifically is AI negative about us, and what is it saying?” This is the question that drives content strategy, PR remediation, and product communications decisions. You might also be wondering:
- “Which of our product categories gets the worst AI treatment?”
- “Is AI negativity on a specific topic driven by one bad source or many?”
- “Are our competitors also receiving negative sentiment on this topic, or is it specific to us?”
Where to Go in Qwairy
Start here: Insights > Sentiment Analysis
Navigate to Insights > Sentiment Analysis — your primary view for topic-level sentiment breakdown.
Apply the Topic or Tag filter to isolate specific areas. Compare the sentiment score for each topic against your overall average. Any topic scoring 15+ points below your average is worth investigating further.
Locate the sources: Overview > GEO Matrix
Cross-reference with Overview > GEO Matrix filtered by the relevant topic or tag.
Identify the specific prompts where negative sentiment is concentrated. The matrix shows you both which prompts and which providers are generating the lowest scores — this tells you whether the negativity is prompt-specific or provider-specific.
Read the negative answers: Visibility > Prompts
Navigate to Visibility > Prompts and open the individual prompts driving the lowest sentiment.
Read the full AI-generated response text for each. The responses will show you the exact language, claims, and framings that are creating the negative signal.
What to Look For
Sentiment Analysis — Topic-Level Breakdown
The Sentiment Analysis view breaks down scores by provider, period, and topic/tag. The topic breakdown is where you find the specific areas of concern.| Element | What it tells you |
|---|---|
| Per-topic sentiment score | The average sentiment for all responses in that topic category |
| Score delta vs. overall | How much worse (or better) a topic performs relative to your brand average |
| Provider breakdown per topic | Whether negative topic sentiment is universal or localized to specific AI models |
| Trend over time per topic | Whether the topic-level negativity is worsening, stable, or recovering |
GEO Matrix — Prompt-Level Identification
The GEO Matrix shows you which specific monitoring prompts are driving the negative signal. Each cell represents a prompt-provider combination, and low-scoring cells in a specific topic row indicate the exact questions where AI is generating negative content about your brand.| Element | What it tells you |
|---|---|
| Low-score rows in topic | The specific user questions that trigger negative AI descriptions |
| Provider columns | Whether the negativity is cross-provider (widespread) or isolated (provider-specific source issue) |
| Competitor row comparison | Whether competitors also score low on these prompts, or whether the negativity is brand-specific |
Pro Tip: Use Tags to group prompts into themes that match your product lines or strategic topics (e.g., “Pricing”, “Security”, “Customer Support”, “Enterprise Features”). This makes topic-level sentiment analysis immediately actionable — each tag becomes a report card for a specific area of your brand.
Filters That Help
| Filter | How to use it for this question |
|---|---|
| Topic / Tag | The primary filter — isolate each area of your brand to find where sentiment drops |
| Provider | Determine whether the negative topic sentiment comes from a specific AI model, which may be sourcing a particular negative article |
| Period | Check whether topic-level negativity correlates with a specific event, a product change, or a media cycle |
How to Interpret the Results
Good result
Topic-level sentiment scores are broadly consistent across your monitored categories — no single topic scores more than 15 points below your overall average. When you read the responses for lower-scoring topics, the negative framing is mild and comparative rather than categorical (e.g., “slightly higher pricing than alternatives” rather than “overpriced and unreliable”). No topic shows a worsening trend over consecutive monitoring periods.Needs attention
One or more topics score below 40 out of 100, with responses that contain explicit negative characterizations: “known for poor customer support”, “frequently cited as difficult to implement”, “has struggled with data accuracy”. The negative sentiment is consistent across multiple providers, suggesting the source material is widely indexed rather than isolated to a single article. A topic’s sentiment has been declining for 3+ consecutive months without stabilizing.Example
Scenario: A sustainability SaaS platform that helps companies track carbon emissions and ESG compliance has an overall sentiment score of 61. They suspect that AI is describing their data accuracy and reporting capabilities negatively, which may be undermining credibility with enterprise procurement teams evaluating ESG tools.
- Open Insights > Sentiment Analysis and apply the Tag filter for “Data Accuracy”. The data-accuracy-tagged prompts produce an average sentiment score of 31, versus the overall average of 61. The gap is 30 points — a significant signal.
- Open Overview > GEO Matrix filtered by Tag = “Data Accuracy”. Three prompts stand out with cross-provider low scores: “most accurate carbon accounting software”, “ESG reporting tools for auditors”, and “sustainability platforms with verified data sources”. All three score below 28 across ChatGPT and Perplexity.
- Open Visibility > Prompts and open the prompt “most accurate carbon accounting software”. Read the full responses. ChatGPT’s response says: “While [Brand] offers a comprehensive dashboard, users have raised concerns about the accuracy of Scope 3 emissions estimates and the lack of third-party data verification. Competitors like [Competitor A] and [Competitor B] are more commonly cited in audit-grade reporting contexts.” Perplexity’s response is similar, citing a sustainability technology review article.
- Open Visibility > Responses and filter by Provider = “Perplexity” and Tag = “Data Accuracy”. Find the responses citing the review article. Note the article URL in the citations panel — it is an 18-month-old comparison that predates the platform’s integration with verified emissions databases.
- Action plan: (a) publish detailed technical documentation on the new verified data pipeline and third-party audit integrations; (b) reach out to the sustainability tech review site for an updated assessment; (c) pursue inclusion in ESG analyst reports and procurement guides that rank for audit-related queries. Set a 90-day monitoring target to bring data accuracy sentiment above 48.
Go Further
Negative sentiment alert dashboard
Build a negative sentiment alert dashboard in Looker Studio to catch topic-level sentiment drops early
Sentiment scoring explained
Read the Sentiment Analysis documentation to understand how negative sentiment is detected and scored by topic
Set up sentiment alerting via API
Use the answers endpoint to build automated alerts that trigger when negative sentiment exceeds your threshold
Related Questions
How has AI sentiment changed after a PR crisis or product launch?
Check whether topic-level negativity correlates with a specific event in your brand timeline.
Which monitoring prompts are most important to track?
Ensure you have the right prompts configured to surface topic-level sentiment signals.
Is AI generating negative sentiment about my brand on specific topics?
When negative sentiment is based on false claims, use the misinformation correction workflow.
What does AI actually say about my brand?
Read the full response narrative before drilling into topic-specific sentiment.