BAX BRAND ATTENTION INDEX Request access
BAX Insights · AI Measurement

What Good AI Attention Looks Like: A Brand Audit Framework

Most brands do not know their AI attention score. Here is how to find out what you are actually getting.

By Tomasz Wilenski  ·  June 18, 2026  ·  8 min read

← All Insights

Step one: Define your query universe

Before you can measure AI attention, you need to know what you are measuring it for. A query universe is the set of prompts that represent how your potential customers actually ask AI about your category.

Most brands start with brand queries: prompts that explicitly mention the brand name. These are the easiest to track and the least strategically useful. A user who already knows your brand name and asks about it is not discovering you through AI. They are verifying you.

The queries that matter are category queries: prompts that describe a need, a problem, or a decision, without mentioning your brand. "What is the best platform for measuring brand visibility in AI?" "Which analytics tools do CMOs use to track attention?" "How do I know if my brand is being cited correctly in AI responses?" These are the queries where AI is acting as the first recommender, and where your presence or absence determines whether you enter the consideration set.

A complete query universe for a brand audit contains a minimum of 50 category queries across three intent types: awareness queries (general category education), consideration queries (comparison and evaluation), and decision queries (specific capability or use case).

Step two: Establish your baseline across platforms

AI responses are not uniform. The same query on ChatGPT produces a different response than the same query on Perplexity, Gemini, or Copilot. Platform-specific behavior reflects differences in training data, retrieval architecture, and response formatting.

A brand audit that covers only one platform is not a brand audit. It is a snapshot of one model's behavior on one day.

Baseline measurement requires running your query universe across a minimum of three to five platforms and recording, for each query and each platform: whether your brand appears, at what position, with what sentiment, cited from what source, and with what accuracy.

This baseline does not need to be large. Fifty queries across five platforms is 250 data points. That is enough to identify patterns, locate outliers, and establish the starting point against which future measurements will be compared.

Step three: Score each appearance

Not all appearances are equal. A brand mentioned first in a direct answer to a decision query on Perplexity, cited from the brand's own domain, in a positive and accurate context, is worth more than the same brand mentioned fourth in a list at the end of an awareness query on a platform with lower intent volume.

Scoring each appearance requires four assessments:

Position quality: where in the response does the mention appear? First paragraph scores highest. Embedded in a list scores lower. Mentioned as a counterexample scores lowest.

Sentiment quality: is the context positive, neutral, or negative, and how cognitively accessible is that context to the user? A positive mention in a dense technical paragraph is worth less than a positive mention in a clear, direct sentence.

Source quality: what domain does the model cite when mentioning your brand? Your own domain is highest. Tier-one industry publications are strong. Aggregators and comparison sites are weak. Competitor domains cited as sources of information about you are a red flag.

Accuracy: does the model describe your brand, product, or capability correctly? Inaccurate positive mentions are not wins. They are liabilities that will be corrected in the next model update, often in ways you cannot predict or control.

Step four: Map your Trust Sphere

Pull the citation domains from all appearances across your query universe. Sort them into three categories: domains you control, domains you influence, and domains that are neutral or hostile to your positioning.

The ratio of controlled to uncontrolled citations is one of the most actionable numbers in a brand audit. If less than 30 percent of citations point to your own domain or content you have directly contributed to, your AI presence is built on ground you do not own. Any model update can shift it.

The Trust Sphere map also identifies your content gaps. If AI consistently cites a competitor's comparison page when mentioning you, it means the model has found that page more authoritative than anything you have published on the same topic. That is a content brief, not just a measurement finding.

Step five: Benchmark against two to three competitors

Run the same query universe for your two or three primary competitors. Record the same data points. Compare.

The comparison reveals four possible competitive positions:

Quadrant 01

You appear more and score higher

You are winning AI attention in this category. Defend and expand.

Quadrant 02

You appear more but score lower

You have volume without quality. Your presence is broad but shallow, often cited from weak sources, often in lower positions. You are vulnerable to a competitor who builds quality rather than quantity.

Quadrant 03

You appear less but score higher

You have a quality foothold without volume. You are cited less often but in better positions, from better sources, with higher accuracy. This is a strong foundation to build on.

Quadrant 04

You appear less and score lower

You are effectively absent from AI-driven consideration in this category. This is the most urgent position and the one that requires immediate content strategy intervention.

Step six: Establish your monitoring cadence

AI models update. Citation patterns shift. A brand that was well-positioned in January may find its position eroded by March if a competitor published authoritative content that the model incorporated in its latest update.

A minimum viable monitoring cadence is monthly measurement of your core query universe across primary platforms, with quarterly deep audits that include Trust Sphere mapping and competitive benchmarking.

The monthly measurement does not need to be comprehensive. Ten to fifteen decision queries across three platforms, scored consistently, will surface material changes before they become strategic problems.

What good looks like

After completing these six steps, you have a baseline. Good AI attention, by the BAX framework, looks like this:

Your brand appears in the first or second position for the majority of your decision queries across primary platforms. The citations underneath your mentions point predominantly to your own domain or to tier-one publications where you have contributed authoritative content. Sentiment is positive in low-cognitive-load contexts. Accuracy is high: the model describes what you actually do, not a version of what you do that reflects outdated or third-party information. Your BAX Index is trending upward across measurement periods.

If your current audit produces a different picture, you now know exactly where to intervene: in which query types, on which platforms, with which content, and in which citation domains.

That is what measurement is for.

Related

What the BAX Index Actually Measures (And Why Everything Else Is Guesswork)

The Source Problem: Why the Websites AI Cites When Mentioning Your Brand Matter

Why Your Brand's Position in an AI Response Matters More Than Its Position on Google

See how your brand performs in AI responses.

BAX measures AI Attention, not just Presence. Request access and run your first audit.