Platform · Sentiment
Understand the tone before the market does
Being cited is not enough: what counts is how the model describes you, and which source suggested it.
The brands we measure every day
The metrics
Your brand sentiment: how AI talks about you, and next to whom
The tone models use to describe you steers real buying decisions. We measure it systematically and comparatively, from the period total down to the sentence that decided the verdict.
-
Your brand sentiment, measured at the source
A summary indicator of how you are perceived, one that comes not from a survey but from the source the market consults every day to decide.
Tone distribution last 30 days71 index- Neutral 78%
- Positive 17%
- Negative 5%
-
Positive, neutral and negative by cluster
The distribution by question family tells you where tone is an asset and where it is a risk: the negative almost always concentrates, it does not spread.
Tone by cluster number of prompts · week 26- Positive
- Neutral
- Negative
- Comparison 216
- How to choose 188
- Usage occasion 110
- Generic 105
- Price 96
-
Every prompt carries every model’s verdict
Estimated popularity and tone, assistant by assistant, on the same question: the most informative case is disagreement, which a cross-model average would erase.
Prompts analysed popularity · verdict- “which is the best in the category” 114K Neutral
- “which one for everyday use” 43K Positive
- “is brand X worth the price” 29K Disagreement
-
What counts is not how they talk about you, but relative to the others
A tone index only makes sense inside a declared competitive set: it is the comparison that says whether you are perceived better or worse than your category.
Tone index vs competitive set
-
From the verdict to the full answer
The model answers side by side, with the passages that drove the tone highlighted and the source that produced them: the diagnosis reaches the sentence.
Tone × cluster by modelUsage Choice Price Assist. ChatGPT 86 74 52 30 Perplexity 78 70 44 26 Gemini 72 61 38 14 Claude 80 66 48 22 -
Drifts are caught before they settle
When tone worsens you trace the cause: which domains feed that judgement and which attributes the brand comes off badly on. The module works in tandem with Fact Checking and External Sources.
Sources behind negative tone mock-up- community-a.com/thread 34 answers Critical
- media-a.com/review 18 answers Lukewarm
- reference.org/guide 9 answers Positive
The tone AI uses to describe you is not a matter of style: it is the sentence your customer reads instead of yours.
Frequently asked questions
Taken verbatim from the monitored prompts. FAQPage schema.
Is the sentiment generated by a model?
Yes, with classification validated on a sample by analysts. The rate of agreement with human review is reported in the platform.
Why is almost everything neutral?
Because that is the prevailing tone of generative answers. The value of the metric is in the two tails, not in the share of neutral.
What do I do about a negative question?
You trace the cited source: in most cases the tone comes from a specific domain, not from the model.
Is it available by language?
Yes, segmented by language and market, with native prompts.
Try it on your own data
Start the 7-day free trial yourself (card required, no call needed): see your real data before choosing a plan.
