Brand Perception
Net Sentiment Score, the attribute matrix, and the negative mentions behind the numbers.
Being mentioned in AI answers is not the same as being recommended. Perception scores how answers talk about your brand: the tone of each mention, the attributes engines associate with you versus competitors, and the specific negative mentions behind a score change. Perception is included in every plan and fills in automatically from your tracked answers.
Net Sentiment Score
Each measurement window gets a Net Sentiment Score (NSS) from -100 to 100, mention-weighted across that window's classified brand mentions. The table shows the score per window with the mention count and the positive and negative split, so a moving score always comes with the volume behind it. A window with no brand mentions has no score; sentiment appears once your brand is mentioned in answers.
Negative mentions
The negative mentions list shows the most recent negative classifications with their evidence: a short excerpt of the answer text that carried the negative tone, per brand and date. When the score drops, this is where you find out why, engine by engine, before deciding what to fix. For a remediation workflow, see Recommendations.
Attribute perception
The attribute matrix shows what engines associate with each brand: one row per attribute (for example "reliable" or "expensive"), one column per tracked brand, with average prominence and mention counts in each cell. A stronger cell means the attribute appears earlier and more often in answers about that brand. Attributes are clustered automatically from what engines actually say, so the vocabulary is the market's, not a fixed taxonomy.
Use the matrix to spot positioning gaps: attributes competitors own that you do not, and negative attributes attached to your brand that content can address.
Alerts
The Sentiment shift alert rule notifies you when NSS moves past your threshold between measurement windows. See Alerts.
Exporting
Perception data is available through the API: sentiment for window aggregates, attributes for the attribute registry, and attribute-mentions for run-level evidence.