What Sentiment Measures
Scrunch tracks how your brand is described in AI-generated responses using Sentiment Average (API field: brand_sentiment_score). This is a quantified 0-100 score, not just a qualitative label: each response mentioning your brand is scored positive = 100, mixed = 50, or negative = 0, and the metric is the average of those scores across all responses that mention the brand.
Looking for the exact Explorer name or API field for a metric? The Scrunch Metrics Reference lists every metric, breakdown and filter with the API field it maps to.
Underneath that average sits the Sentiment band (API field: sentiment_band), which classifies each individual response into one of four categories:
Positive: Language is favorable or endorsing
Mixed: Includes both pros/cons or neutral observations
Negative: Highlights drawbacks or positions your brand as less desirable than another option
Unclear: The response mentions your brand but sentiment can't be reliably determined (this is the 'none' value of the Sentiment band)
This metric is powered by a machine learning (ML) model trained to evaluate tone and positioning across thousands of AI outputs.
How Negative Sentiment Is Triggered
Unlike traditional social listening tools, Scrunch's sentiment model isn't just looking for negative keywords. Instead, it captures relative sentiment:
The most common case: your brand is mentioned alongside competitors, and the AI recommends a different option.
Example: "For your use case, definitely go with [Competitor]," while your brand is mentioned but not recommended.
It is rare for an AI to outright say "Brand X is bad", since most are tuned to maintain a neutral or optimistic tone.
This means a response landing in the Negative Sentiment band usually signals competitive disadvantage in positioning, not explicit criticism, and pulls your overall Sentiment Average down.
Current Limitations
There isn't a one-click list of "here are your negative mentions." You explore sentiment by building a view in Explorer.
In Explorer, add Sentiment Average as a metric and filter by Sentiment band, then break the results down by Prompt or Citation URL to see which prompts or cited sources are pulling your score down. Explorer supports up to 5 metrics, 3 breakdowns, and 8 filters in a single view.
Sentiment is tracked at the response and brand level, not per citation. There is no separate sentiment value attached to an individual citation, so you can only cross-reference the Sentiment band of the response a citation appeared in, not a sentiment score for the citation itself.
For more granular tracking, you can use the API or the Data Studio connector to slice sentiment data down to the prompt or response level.
What's Coming Next
We're actively working on:
Expanded Insights that will highlight sentiment-related opportunities, including where competitor mentions are pushing your Sentiment Average down.
How to Use Sentiment Data Today
Identify weak spots: In Explorer, filter by Sentiment band = Negative and note if a competitor is favored in those same responses. Leverage the 'Brand Protection' category in the Insights tab for help identifying these prompts and responses.
Prioritize fixes: Focus on the prompts and cited sources driving down your Sentiment Average.
Take action:
Strengthen content around those topics.
Consider PR or backlink strategies to shift how third-party sources present your brand.
Track your Sentiment Average over time as new data is collected.
Key Takeaway
Scrunch's Sentiment Average helps you understand not just if you're being mentioned, but how you're being positioned. This metric provides early signals of competitive disadvantage so you can act before it impacts perception more broadly.