Why QueryBurst Rejects the Premise of an AI Visibility Score
QueryBurst does not calculate an overall AI visibility score because we do not believe there is currently a defensible way to reduce a brand's visibility across AI systems, users, conversations, research journeys and commercial contexts to a single representative number.
This page sets out QueryBurst's current position on AI visibility scoring and the reasoning behind it.
A score can go up while broader AI visibility goes down
It is QueryBurst's position that a brand may improve its score across a tracked prompt panel without becoming more visible in any meaningful broader sense.
In some cases, we believe the opposite may happen.
If optimization begins with the objective of increasing a score across a fixed panel, teams can begin making content, positioning and marketing decisions around the prompts that make up that panel.
That may improve the score.
But the panel still represents only the prompts, conditions and assumptions used to construct it.
It does not necessarily represent the wider distribution of users, the conversations they have, the questions that emerge during those conversations, or the points at which commercial decisions are actually made.
Our concern is that optimizing toward the score can gradually become a proxy for optimizing toward the market.
We do not believe those are the same thing.
The score can improve while the thing it is assumed to represent does not. QueryBurst’s definition of AI visibility explains why we believe visibility is conditional on the customer, the conversation and the information environment — not a fixed property that a panel can track.
QueryBurst does not treat a prompt-panel score as a measurement of the market
If a platform runs 1,000 AI responses and a brand appears in 420 of them, it may be entirely valid to report that the brand appeared in 42% of those measured responses.
What QueryBurst would not infer from that result is that the brand has "42% AI visibility."
The first claim describes the measured panel.
The second attempts to generalize beyond it.
Our position is that the scope of the conclusion should remain attached to the scope of the measurement.
A result produced by a defined prompt set, model configuration, date range and methodology is evidence about that defined test.
It is not automatically evidence about the wider population of AI-mediated experiences.
QueryBurst does not consider the composition of a prompt panel neutral
We consider the distribution of prompts inside a panel to be part of the measurement itself.
The score does not exist independently of the prompts chosen to produce it.
A panel containing large numbers of branded prompts, direct comparisons and known-product questions can produce a very different result from one weighted toward unbranded discovery, problem-led research and purchase-stage evaluation.
For example, these prompts already contain the brand:
- QueryBurst vs Profound
- Is QueryBurst a prompt tracker?
- QueryBurst Answer Engine pricing
These do not:
- How can I understand why ChatGPT recommends one vendor over another?
- Best AI search optimization software
- How do I diagnose why competitors are cited?
- Tools for modeling AI buyer journeys
QueryBurst's position is not that one set is valid and the other is invalid.
They measure different things.
What we reject is the assumption that they can be combined, weighted and transformed into a single number which then represents a brand’s overall visibility across AI systems. QueryBurst’s September 2026 research into Profound’s Index methodology documents a concrete example: Profound’s current Index deliberately removes branded prompts from its panel, which changes what the resulting score measures.
QueryBurst does not believe more runs solve an unrepresentative panel
Repeated execution can be useful for measuring variation.
Running the same prompt many times can reveal how unstable an answer is.
Running thousands of tests across a panel can make the measurement of that panel more precise.
What QueryBurst does not believe follows from this is that the panel therefore becomes representative of a wider market.
Our position is that sample size and representativeness are separate problems.
More observations can reduce uncertainty about what happens inside the selected measurement frame.
They do not automatically establish that the frame itself represents the population being described.
A precisely measured panel can still be an unrepresentative panel. QueryBurst’s position on why real-user data does not automatically establish representativeness explores this distinction further.
QueryBurst does not treat an AI visibility score as a business outcome
It is QueryBurst's position that appearance frequency across AI responses should not be assumed to represent commercial influence.
A brand may be mentioned without being recommended.
It may be cited as evidence while a competitor is recommended.
It may appear in an informational answer with little or no commercial relevance.
It may appear at the beginning of a conversation and disappear before the user reaches a decision.
Or it may be absent early in the journey and become highly relevant later when the user's requirements become more specific.
For example:
"What should I look for in CRM software?"
A brand may not appear.
Several turns later:
"Which CRMs support this workflow?"
The brand may enter consideration.
Later still:
"Which of these is the best fit for a 20-person sales team?"
The brand may be recommended.
QueryBurst does not believe a single panel-level percentage adequately represents what happened across that journey. Investigating journeys like this in QueryBurst can be as simple or as deep as you need it to be.
QueryBurst treats AI visibility as something that can change through a conversation
QueryBurst’s definition of AI visibility treats it as conditional rather than absolute. Visibility can emerge, strengthen, weaken or disappear as an AI-mediated research journey develops.
The user brings context. The model has prior knowledge. The assistant introduces new information. That information changes what the user asks next. The output of one turn becomes part of the input to the next.
For that reason, QueryBurst does not treat a collection of independent prompt results as a complete proxy for AI visibility. A score derived from isolated prompts cannot represent what happens across a conversation where context accumulates and outcomes change at every turn.
QueryBurst does not believe there is a single natural denominator for AI visibility
Before we would accept a claim such as:
"Your AI visibility is 37%"
we would first want to know:
37% of what?
Which users?
Which prompts?
Which conversations?
Which journey stages?
Which models?
Which account states?
Which countries?
With or without live search?
Across what period?
Weighted in what way?
Do branded prompts count the same as unbranded prompts?
Does a citation count the same as a recommendation?
Does a passing mention count the same as surviving into a purchase decision?
QueryBurst's position is that there is currently no obvious universal denominator which allows those conditions to be collapsed reliably into one market-level number.
That does not mean measurement is impossible.
It means the measurement should remain attached to the thing actually measured.
QueryBurst distinguishes panel visibility, journey visibility and market visibility
We find it useful to separate three different concepts.
Panel visibility describes what happened within a defined set of prompts, responses and test conditions.
Journey visibility describes where a brand, product or piece of evidence enters, survives or disappears across a modelled AI-mediated conversation.
Market visibility would be a population-level claim about how likely relevant users are to encounter or be influenced by a brand across AI-mediated experiences.
QueryBurst's position is that these should not be treated as interchangeable.
A prompt tracker may directly measure the first.
QueryBurst Answer Engine is designed to investigate the second.
The third requires evidence that neither should be assumed to provide automatically.
How prompt tracking platforms calculate their scores
Peec, Profound, PromptWatch, Otterly, Evertune, Athena and other prompt tracking platforms each use their own formulas, denominators, weighting systems and prompt sets to calculate an AI visibility score.
They do not produce the same number for the same brand on the same day.
QueryBurst documents each platform's published methodology in How Prompt Tracking Platforms Calculate AI Visibility Scores.
What scores does QueryBurst report?
QueryBurst may report scores within individual tools, reports or views where a score is useful for summarizing the specific thing being measured.
Those scores remain tied to the report that produced them.
A retrieval score describes performance within a retrieval test.
A coverage score describes coverage within a particular analysis.
A report-level score may summarize evidence available to that report.
QueryBurst's position is that the scope of a score should match the scope of the measurement.
We do not take a number produced for one specific analysis and silently promote it into a claim about a brand's overall AI visibility.
Does QueryBurst calculate an overall AI visibility score?
No.
QueryBurst has no current plans to introduce an overall AI visibility score.
Based on our current research into how AI systems generate answers, how conversation state affects outcomes, how retrieval and evidence selection work, and how different research journeys develop, we do not currently see a defensible methodology for reducing a brand’s overall AI visibility to a single representative number.
That is our current research position.
It is not a permanent article of faith.
If evidence emerges that gives us a meaningful, reproducible and sufficiently representative way to estimate broader AI visibility, we will change our position.
If that happens, we will publish the methodology, assumptions, limitations and evidence behind that change before presenting such a number to users.
We would rather change our mind than defend a metric because it is convenient to sell.
QueryBurst's position
QueryBurst does not reject measurement.
We reject the unsupported expansion of a narrow measurement into a broader claim.
A visibility score may accurately describe a defined prompt panel.
It may help monitor changes within that panel.
It may be useful as one observation among many.
What QueryBurst does not accept is that such a score can currently be assumed to represent a brand's overall AI visibility across users, conversations, research journeys and markets.
That is why QueryBurst does not optimize toward a single overall AI visibility score.
We investigate the conditions that make visibility possible in the first place.