How Peec, Profound, Evertune & Other Prompt Tracking Platforms Calculate AI Visibility Scores
QueryBurst does not calculate an AI visibility score. Our position on why is set out in Why QueryBurst Rejects the Premise of an AI Visibility Score.
This page documents how prompt tracking platforms calculate theirs, based on their own published documentation as of September 2026.
What prompt tracking platforms have in common
The platforms listed below all share a common approach: run a defined set of prompts against one or more AI models, detect whether a brand appears in the response, and aggregate those observations into a score.
The differences are in how they aggregate.
Some count mentions. Some weight by position. Some adjust for sentiment. Some exclude responses where no brand appeared. Some composite multiple signals into a single index.
All describe the result as an AI visibility score. They should not be expected to produce the same number for the same brand on the same day.
QueryBurst's position is that this is not a calibration problem. It is a consequence of the fact that "AI visibility score" does not refer to a single agreed measurement. Each score reflects the vendor's own definition of what AI visibility means, operationalized through their own methodology.
Published methodologies compared
The following table summarises how several prompt tracking platforms calculate their headline score, based on their own published documentation as of September 2026.
| Platform | Headline metric | Published formula | Denominator | Position weighting | Source |
|---|---|---|---|---|---|
| Peec | Visibility Score | Responses mentioning brand ÷ total responses × 100 | All tracked responses | No (reported separately as Position) | docs.peec.ai |
| Profound | Visibility Score | Responses including brand ÷ responses containing at least one brand × 100 | Only responses with ≥1 brand mention | No (reported separately as Average Position) | help.tryprofound.com |
| PromptWatch | Visibility Score | Per response: Position + Context − Competitor weight. Overall: mean of all response scores, absences counted as 0 | All tracked responses | Yes (lead recommendation scores highest, lower placements score less) | promptwatch.com |
| Otterly | Brand Visibility Index | Two-dimensional index. x-axis: Brand Coverage (% of responses brand appears in). y-axis: Likelihood to Buy, derived from average position using Likelihood = 10 + ((5 − avgPosition) / 4) × 90 | All tracked responses (for coverage component) | Yes (position-derived “Likelihood to Buy” with 10% floor) | help.otterly.ai |
| Evertune | Visibility Score / AI Brand Score | Visibility Score = % of responses mentioning brand. AI Brand Score = Visibility Score weighted by position. Pos 1 = 100%, Pos 2 = 90%, Pos 3 = 81%, each subsequent −10% of previous | All tracked responses (each prompt sampled 100×) | Yes (geometric decay: each position 90% of previous) | evertune.ai |
| QueryBurst | Does not report an AI visibility score | N/A | N/A | N/A | Position statement |
The denominator is not consistent across platforms
Peec divides by all tracked responses. Profound divides by responses containing at least one brand, which excludes responses where no brand appeared.
This means Profound's denominator is structurally smaller. Where some responses contain no brand at all, the same underlying response set can produce a higher percentage under Profound's formula than under Peec's.
Both call the result a visibility score.
Position weighting varies or is absent
Peec and Profound treat all mentions equally regardless of where the brand appeared in the response. A passing mention at the end of a list counts the same as a lead recommendation.
PromptWatch, Otterly and Evertune weight by position, sentiment, or both. Two platforms can observe the same set of responses and produce materially different scores depending on where the brand was mentioned and how many competitors shared the answer.
These are not errors. They are design decisions about what "AI visibility" means.
The same brand can have different scores on the same day
Because each platform uses a different formula, denominator, weighting system, prompt set, execution conditions, run count and aggregation method, two platforms tracking the same brand on the same day should not be expected to produce the same score.
PromptWatch's own documentation acknowledges this directly: "Two tools will return different visibility scores for the same brand on the same day, and neither is lying. A composite score is a weighting decision."
QueryBurst's position is that the scores are vendor-specific measurements, not vendor-neutral facts about a brand's AI visibility.
QueryBurst's position
QueryBurst does not dispute the arithmetic behind any of these methods.
Our disagreement is with the interpretation placed on top of it.
A score derived from a tracked prompt panel may accurately describe that panel. We do not believe it should automatically be promoted into a claim about a brand's overall visibility across AI systems, its share of a market, or its commercial influence.
Each vendor's score reflects their own answer to the question "what does AI visibility mean?" QueryBurst's answer to that question is set out in How QueryBurst Defines AI Visibility.