Sentiment

How favourably AI answers describe your brand, on a 0–100 scale built from per-answer verdicts.

Appearing in an answer is not automatically good. Sentiment measures how your brand is described when it is named — recommended, listed neutrally, or warned about.

Verdicts

Each answer that mentions your brand receives one verdict about that brand:

VerdictMeaning
PositiveThe brand is described favourably — recommended, praised, presented as a strong choice
NeutralThe brand is named without a judgement — listed, compared, stated as a fact
MixedThe same answer says both favourable and unfavourable things
NegativeThe brand is described unfavourably — criticised, or presented as the option to avoid

The verdict is about your brand specifically, not about the mood of the answer as a whole. An answer that is enthusiastic about a competitor and dismissive of you is negative for you.

The 0–100 score

Verdicts across the period are combined into a single score:

Sentiment = ((positive answers − negative answers) / scored answers + 1) / 2 × 100

Read it against the midpoint:

  • 50 — balanced, or entirely neutral coverage
  • above 50 — favourable answers outweigh unfavourable ones
  • below 50 — unfavourable answers outweigh favourable ones
  • 100 / 0 — every scored answer was positive / negative

Neutral and mixed answers count in the denominator but not in the numerator: they pull the score toward 50 rather than up or down. A brand described neutrally everywhere scores 50, which is the honest reading — assistants mention it without taking a side.

The score is shown only when at least one answer mentioning the brand was scored in the period. With no mentions there is nothing to score, and the dashboard shows no value rather than a misleading 50.

Worked example

In the period, 20 answers mention your brand: 9 positive, 8 neutral, 1 mixed, 2 negative.

Sentiment = ((9 − 2) / 20 + 1) / 2 × 100 = 67.5 → 68

Why sentiment matters

  • It qualifies visibility. Rising visibility with falling sentiment means assistants are increasingly naming you as the option to avoid — a worse outcome than not appearing.
  • It points at source material. Assistants describe you the way their sources do. A drop usually traces back to what is being published about you, which is why it is read alongside Citations & Sources.
  • It is comparable to competitors, scored the same way over the same answers.

Reading the number carefully

  • Small samples swing hard. With five scored answers, one verdict moves the score by ten points. Watch the trend line rather than the daily value.
  • A high score on very low visibility describes a handful of answers, not your market position.
  • Mixed is information, not a failure. A cluster of mixed verdicts usually means a real, specific reservation is being repeated about your product.

Updated 2026-08-03