How QueryBurst Defines AI Visibility

AI visibility is the extent to which a brand can enter and influence AI-mediated customer research across the people, conversation states, questions and research paths relevant to its business.

It is not simply the percentage of generated answers in which a brand is mentioned, cited or recommended.

AI visibility is conditional.

The same brand can be highly visible in one customer journey and absent from another. The same question can produce materially different research paths depending on who is asking, what the model already knows, what has happened earlier in the conversation, which constraints have been introduced, what evidence is available, and what the system retrieves and uses at that moment.

That means there is no single context-free state called “AI visibility” that can be observed directly.

Platforms which rely on prompt panels sample a selected set of those conditions. If a brand appears in 37 of 100 collected answers, then 37% can accurately describe its presence within that panel under those execution conditions. Whether those 100 observations represent a wider customer population or market is a separate methodological question.

QueryBurst does not track prompts, does not report an AI visibility score, and should not be evaluated as a prompt tracking platform.

Instead, QueryBurst approaches AI visibility the same way marketers have approached customer journeys for decades: identify the people you want to reach, map the plausible paths they may take, and understand what influences them along the way.

The difference is that those journeys now unfold through conversations with AI. Context accumulates, questions change, research branches, and different evidence can enter at different points.

QueryBurst provides a workspace for investigating those journeys and improving the information environment that shapes them. It is designed to be as simple or complex as you need it to be. Some users may want to investigate a single persona and conversation. Others may want to explore multiple personas, branch conversations, and inspect the research and evidence behind every answer. Everything is there when you need it. And gets out of your way when you don’t.

We believe improved AI visibility is an outcome, not a score.

Your customers are having conversations with AI. QueryBurst lets you step inside them.

AI visibility is conditional, not absolute

A brand’s visibility in any given AI response depends on the conditions surrounding that response.

Who is the customer? What do they already know? What have they asked before? What constraints have they introduced? What has the model already told them? What information is available at that moment? What does the system retrieve, and what does it do with what it finds?

Change any of those conditions and the outcome can change — even for the same underlying question.

QueryBurst’s position is that this conditionality is not a complication to be averaged away. It is the central characteristic of AI visibility. We set out our argument for this in our opinion piece ChatGPT is ‘Invisible’ to Your Prompt Tracker.

A market is not a collection of interchangeable prompts. It contains people moving through different conversational states.

AI visibility begins before the user asks a question

An AI system does not begin every conversation with a blank slate.

Its existing knowledge and associations can influence which brands, products, categories, attributes and relationships are already available before retrieval begins.

A brand can therefore already be available to influence the research path before it appears in a generated answer.

If a brand is strongly associated with a category, problem, feature or use case, that prior knowledge may influence what the system investigates next. If the association is weak, incorrect or absent, that can influence the conversation too.

QueryBurst treats this prior state as part of the AI visibility problem rather than starting measurement only after a prompt has been submitted.

AI visibility changes as conversations develop

The current user message is only one part of a multi-turn conversation.

Previous questions matter. Previous AI responses matter. Products already discussed matter. Requirements, objections, rejected options and new information introduced during the conversation can all change what becomes relevant at the next turn.

The order matters too.

The same question asked at the beginning of a conversation can exist in a materially different context when asked after several rounds of research.

QueryBurst therefore treats conversation state as something that develops over time rather than reducing each turn to an isolated prompt.

The final answer is only one observable stage of AI visibility

The final answer is the part of an AI interaction the user can see.

It is not the only part that can determine whether a brand appears or influences the result.

AI systems may generate additional research queries, retrieve sources, extract passages, compare evidence, resolve competing claims and determine which information is useful enough to support an answer.

A brand can enter this process without appearing in the final response.

A page can be retrieved without being used.

Evidence can influence an answer without receiving the visible citation.

A source can be cited without its brand being recommended.

QueryBurst treats these as different stages of the visibility problem rather than collapsing them into whether a brand name appeared in the final output.

The final response is evidence of what happened. It is not a complete record of why it happened.

A prompt panel samples AI visibility; it does not define it

A prompt panel can be measured accurately.

If a brand appears in 37 of 100 collected answers, then 37% accurately describes its presence within that panel under those execution conditions.

What it does not establish is that the brand has 37% AI visibility in a market.

The panel is a constructed sample of customer states. Its result depends on which questions were included, which were excluded, how they were weighted, which people or personas they represent, what conversational state preceded them, which systems were queried and under what execution conditions.

Change the panel and you can change the score without anything about the brand or market changing.

The reverse is also true. A brand that optimises for a predefined panel of prompts may reach a perfect score without improving its underlying visibility across the customer journeys that actually matter. The score can improve without the conditions determining wider customer outcomes improving with it.

Real customer demand also contains branded questions, comparisons, follow-ups, objections, rejected recommendations, changing constraints and accumulated context.

A benchmark may deliberately exclude some of those behaviours. That can be a valid methodological choice. But the resulting score then describes performance within that benchmark rather than visibility across the market as a whole.

QueryBurst’s September 2026 research into Profound’s conversation dataset, Prompt Volumes and Index methodology provides a concrete example. Profound’s current Index deliberately filters toward open-ended, brand-agnostic prompts and removes prompts that name specific brands. That creates a standardised brand-agnostic benchmark, but it does not reproduce the complete distribution of branded and unbranded buyer conversations.

A prompt tracker can measure its panel accurately. That does not make the panel a measurement of the market.

QueryBurst’s positions on AI visibility scores, prompt tracking and real-user data representativeness are set out in detail elsewhere in this hub.

Mentions, citations and recommendations are observations from AI visibility

A mention tells us the brand appeared.

A citation tells us a source was surfaced visibly.

A recommendation tells us the brand survived far enough through the process to be suggested in that response.

They are observations from AI visibility, not the underlying phenomenon itself.

None independently tells us why it happened, whether it would happen for a different customer state, or how representative that outcome is of a wider market.

The more useful question is what caused the brand or evidence to enter the conversation, survive the research process and influence the resulting answer.

That distinction matters because the same visible outcome can be produced through very different paths, and the same underlying visibility can produce different outputs as context changes.

QueryBurst’s detailed breakdown of sources, citations, mentions and recommendations explains each stage and why different failures at different stages require different fixes.

QueryBurst starts with the customer journey, not the prompt panel

QueryBurst users define the people and research journeys they want to investigate.

The persona establishes who the customer is. The conversation records what has happened to them.

Those journeys can include different needs, constraints, levels of knowledge, commercial intent, objections, follow-up questions and decision paths.

QueryBurst then runs them as stateful conversations in which previous turns and AI responses remain part of the context for what happens next.

This makes it possible to investigate visibility as something that develops across a conversation rather than as a percentage derived from a predefined collection of independent prompts.

QueryBurst does not claim to measure the market either. The full space of customer states that constitute AI visibility is not directly observable as a complete market. What QueryBurst provides is an open-ended investigation space — users can explore as many people, conversation paths and contexts as matter to their business, without being constrained to a fixed panel. The result is not a market-level measurement. It is an understanding of the conditions that produce specific outcomes.

Using QueryBurst starts with a persona and a conversation. The depth of investigation scales with what you need.

QueryBurst makes the answer-generation process inspectable

Production AI systems do not expose their complete internal answer-generation process to third parties.

QueryBurst does not claim that its workspace exposes the proprietary internal architecture of production AI systems.

QueryBurst constructs an inspectable operational model of the stages that matter for investigation: prior knowledge, research queries, retrieval, source selection, evidence and answer generation. How we research and build those models is documented separately.

Within a QueryBurst run, those stages are inspectable.

That lets users investigate questions that cannot be answered by collecting final outputs alone: what the model’s operational behaviour indicates about prior brand knowledge, what information the system went looking for, which sources entered the research process, which evidence survived within the modelled research path, where a website dropped out, and how the available evidence shaped the answer.

AI visibility starts with the information you control

Your website is the part of the information environment you control directly.

You decide what it says about your business, products, services, customers, capabilities, limitations, pricing, evidence and competitive position.

You decide which questions it answers and which information is missing.

You also decide how clearly claims are stated, whether they are consistent across the site, and whether the evidence required to support them exists.

Site Intelligence provides tools for investigating and improving this owned information layer because weaknesses here can affect both the visible and hidden stages of AI answer generation.

This is what we mean when we say AI visibility starts at home.

AI visibility also depends on what the wider web says

Your website cannot establish every claim about your brand on its own.

Independent coverage, customer reviews, expert opinion, links, third-party data, reputation and other external evidence can reinforce, qualify or contradict the information you publish.

QueryBurst is designed around the part of AI visibility you can directly investigate and improve, but stronger owned information does not remove the need for external corroboration.

The strongest position is created when what your website says is clear, well-supported and reinforced by evidence beyond your own domain.

The goal is to improve AI visibility, not optimise a score

QueryBurst is designed to help users investigate where visibility is gained or lost and improve the conditions that influence it.

That can mean clarifying what a website says, correcting inconsistent claims, filling information gaps, strengthening evidence, improving how important information is surfaced, or understanding why a brand enters one research path but not another.

The objective is not to optimise an arbitrary collection of individual answers.

It is to improve the underlying information environment so that your brand has a stronger chance of entering and influencing the conversations that matter to your customers.

AI visibility is not a number attached to a brand. It is an emergent property of the relationship between the customer, the conversation, the model’s existing knowledge, the available evidence and the system’s research process.

QueryBurst is designed to investigate and improve those conditions.

Over time, the aim is not simply to become another possible answer.

It is to become the obvious answer.