Browse articles & contentsSearch all help

In this article

Sentiment overviewHow to read the distribution and the topic mixHow to find what’s driving itWhat to do about a negative trend

Getting started

Sign inYour free auditAdding a brandBrowser extension

Understanding your results

OverviewAI visibilitySentimentSourcesCompetitorsMarketSite auditPage auditTrackersSSPSSSPS+PERFYour PDF reportThe opening slidesThe site audit slidesYour business and your marketHow you appear in AI answersSources, trackers and competitorsThe plan at the end

Taking action

RecommendationsContextTechnicalContentContent briefsFunnels

Setup

Prompts and topicsAI enginesBrandUsage & Scan scheduleIntegrations

Account

Account settingsTeam membersBillingHistory & AccessDanger zone

AI agents (MCP)

AI agents (MCP)

Help Understanding your results

Sentiment

Use the Sentiment screen to see how AI platforms describe your brand when they mention it.

Sentiment overview

Being mentioned isn’t the same as being recommended. Sentiment looks only at the answers where you appear and scores the tone of what was said.

The period control at the top switches between 7, 30 and 90 days.

How to read the distribution and the topic mix

Sentiment distribution shows the split across your latest scan.

By topic puts the same latest scan next to itself, one bar per topic, split into the five tone levels. An average alone hides the shape: “mostly neutral” and “half praise, half complaints” average to the same number and mean very different things.

Each says plainly when there’s nothing to show yet, rather than drawing an empty chart.

This screen is the latest scan. Whether the tone is moving — scan over scan, overall and per topic — lives on the Trackers screen, where every dynamic in the product now lives.

How to find what’s driving it

Two breakdowns narrow it down further:

  • Most positive prompts and Most negative prompts — ranked by average score across the period, with how many runs each is based on.
  • By AI platform — average score and run count per engine.

Start with the negative prompts. A single prompt scoring badly across every engine usually points at one source saying something unflattering; the same prompt scoring badly on one engine only points at that engine’s sources.

Good to know Every number here is based on a run count, and it’s shown next to the score. A topic with three runs behind it can swing wildly between scans — check the count before reacting to a move.

What to do about a negative trend

Sentiment isn’t something you edit. It changes when what’s published about you changes.

The route is the Content queue: cards raised for weak or negative tone on a theme come with a brief for what to publish about it. See Content.