Understanding Prompt Lab metrics

What mention rate, position, sentiment, and volume mean — and how to read them without fooling yourself.

Updated 2026-10-12

Every prompt row in Prompt Lab summarizes many runs across providers. This article explains what each number means and the common misreadings to avoid.

Mention rate

The fraction of providers whose latest answer mentions your brand — "1/1" means the one enabled provider mentioned you in its latest run. At the top of the page, the aggregate mention rate covers all tracked prompts.

Maya detects mentions of your brand name, its spelling variants, and your domain. A mention is counted even when the brand is not linked.

Position

Where your brand sits in the answer when it is mentioned. "Best #1" means the assistant's answer treated you as its first recommendation. Position matters more than raw mentions: in a list of eight hotels, being first and being last are very different outcomes, and AI answers are consumed top-down even more than search results pages.

The header's average position aggregates only the answers where you appeared.

Sentiment

When your brand appears in an answer, Maya scores how the answer talks about it, from clearly negative to clearly positive, and shows a label per prompt:

  • positive — the answer recommends or praises your brand.
  • neutral — factual coverage without a lean, or your brand did not appear at all. Sentiment can only be measured about you when the answer actually discusses you; prompts where you are absent are shown as neutral, not negative.
  • negative — the answer criticizes your brand or recommends against it.

A sarcasm check and an answer-quality check guard the score against unreliable answers.

Common misreading: a negative label on a discovery prompt where you never appear. That is not "the AI dislikes us" — being absent is a visibility problem, not a reputation problem. (Older versions of Maya could show "negative" in that case; current behavior shows neutral.)

Volume

An estimate of how often questions like this prompt are actually asked, bucketed from Very Low to Very High. Use it to prioritize: a #1 position on a Very High volume prompt is worth more than five #1s on Very Low prompts.

Trend

The sparkline tracks your mention/position movement over recent runs. Flat lines on an unmentioned prompt are normal — the interesting moment is when a competitor's position shifts or you first enter the answer.

Reading a single answer

Open a prompt to see the full answer per provider, with:

  • Confidence — how certain the mention detection is.
  • Sentiment highlighting — turn on "Show Sentiment" to see which sentences drove the score.
  • Top competitors — the brands and domains that appeared in this prompt's answers, with how often they showed up.

Rules of thumb

  1. Watch position within answered prompts and mention rate across prompts separately — mixing them hides movement.
  2. Compare like with like: providers differ systematically; a drop isolated to one provider is a provider story, not a brand story.
  3. Give changes a week. Single-day swings are often answer randomness, not real shifts.

Related articles