Sentiment, explained

Getting mentioned by an AI assistant isn't automatically good news. If the AI says your brand has "a steep learning curve" or is "not recommended for beginners," that's a mention that could actually be hurting you. Sentiment is how you tell the difference.

How sentiment is measured

Every individual brand mention gets its own sentiment reading: positive, neutral, or negative, based on the specific context immediately around your brand's name in that response, plus a confidence score. This is measured per mention, not per whole AI answer, which means a single response can genuinely praise your brand while criticizing a competitor right next to it, or vice versa, and both readings come through correctly rather than getting blended into one number for the whole response.

Those three are the only values. There's no fourth "unrated" or "mixed" bucket: a mention whose language cuts both ways is resolved to neutral, so neutral covers both a purely factual mention and a genuinely two-sided one. The confidence score is where that distinction shows up, which is why it's worth reading alongside the label rather than on its own.

Very occasionally the model returns something that isn't one of the three at all. In that case no sentiment is recorded for the mention, and it's left out of your percentages entirely instead of being quietly counted as neutral.

How your brand-level sentiment percentage is calculated

To get your overall "% positive" figure, each prompt run is first collapsed to a single best sentiment result for that run, the same collapsing logic used for scoring. Then:

% positive = positive-labeled runs / all sentiment-labeled runs (in your selected window)

A more detailed breakdown is also available at the per-prompt level, showing the full positive/neutral/negative split for that specific prompt, along with a "dominant sentiment" label, whichever bucket has the most mentions for that prompt.

Sentiment Check is a different tool, a different number

They measure different things:

• Dashboard sentiment (described above) is ongoing, tied to your actual tracked prompts and runs, and reflects real per-mention tone over time.
• Sentiment Check is a one-off, on-demand check of how AI generally perceives your brand as a whole. It scores sub-dimensions like trust, innovation, value, and customer service on a 0-100 scale each, cites the real sources it drew from along with each source's individual stance, and adds a margin-of-error range around its headline score based on how many sources it found.