Sentiment

Visibility tells you AI engines name you. Sentiment tells you whether that’s helping or hurting. The Sentiment page reads every claim AI makes about your brand — and your competitors — and rolls it up into a score you can track, broken down by the features buyers care about. For the precise formulas behind every number here, see the Metrics reference.

How sentiment is scored

Each claim an answer makes about a brand is labelled:

• Endorsed — it helps the brand (“Acme is easy to adopt”).
• Neutral — it merely describes (“Acme is a CRM”).
• Undermined — it works against the brand (“Acme lacks enterprise controls”).

Those labels roll up into net sentiment (NS) — a single score from −1 to +1 that captures whether a brand’s narrative leans positive or negative. NS is shown with a credible interval: a band that widens when there’s little evidence and tightens as more answers accumulate, so you can tell a confident score from a noisy one. (The dashboard’s Sentiment tile shows the same idea on a simpler 0–100 scale — see the Dashboard.)

The sentiment trend

The header card tracks your brand’s net sentiment over time, with the credible-interval band shaded around it. While there’s too little evidence to call a trend, it shows a “baseline forming” state rather than a misleading line. Three tiles summarise the window:

• NS — your overall net sentiment, with its credible interval.
• Evidence — the number of feature-level sentiment events behind the score.
• Answers — how many captured answers contributed.

A plain-language headline reads the trend for you — whether you’re climbing, holding steady, or in the red — and calls out your strongest and softest features.

Features

Switch between two views with the List / Heat-map toggle.

List view

The features table is the heart of the page. Each row is a feature — an attribute or capability AI discusses, like pricing, support, or integrations — with:

• Your brand — your sentiment for that feature, as a pill (positive, negative, neutral, mixed, or no coverage) with the evidence count behind it.
• Market benchmark — the same read across the competing brands, so you can see the category norm.
• Takeaway — a plain-language verdict: a differentiator where you beat the market, an opportunity or category pain point where everyone struggles, or missing coverage where AI doesn’t yet discuss you on that feature.
• Evidence — the volume of sentiment events behind your score versus the field.

Sort by any column to find where you lead, where you trail, and where the conversation simply hasn’t formed yet.

Heat-map view

The heat-map lays features across the top and brands down the side, colouring each cell by net sentiment — green where a brand is endorsed on a feature, red where it’s undermined, muted where there’s no signal. It’s the fastest way to spot patterns: a column that’s red for everyone is a category-wide weakness; a cell where you’re green and competitors are red is a story worth amplifying.

The evidence behind a score

Click any feature — or any heat-map cell — to open its detail. Sentiment is never a black box: the detail view shows the endorsed / neutral / undermined split, a brand-by-brand breakdown for the feature, and the actual answer excerpts that produced each label. That’s how you go from “AI is lukewarm on our support” to the specific sentences buyers are being shown — and decide what to do about them.

When sentiment appears

Sentiment is produced by a language-model pass over your captured answers, so it lands shortly after your prompts run rather than instantly. New projects show a “baseline forming” or empty state until enough answers have been labelled. Once they have, the page fills in on its own.