The AI Visibility Measurement Framework
A lot of brands working on AI visibility are measuring the wrong things, in the wrong order, with no clear idea of what they are trying to fix. This chapter sets out the framework that makes measurement actually useful.
Why measurement without structure fails
A number on its own tells you nothing
If you have ever looked at an AI Share of Voice score and had no idea what to do next, you already know the problem. The number tells you where you are. It does not tell you why you are there, what is causing it, or what would actually move it.
AI brand visibility is not one single thing. It is a set of conditions, each of which can go wrong in different ways. A brand can show up often in AI responses but be described badly. It can be described well but connected to the wrong topics. It can be connected to the right topics but missing from live search results entirely. Each of these looks different in the data and needs a different fix.
A framework does not make things more complicated. It makes action possible. Without one, you are looking at symptoms with no way to find the cause.
Most tools give you a Share of Voice score and leave it there. That is like a doctor telling you your temperature is high and sending you home. The reading is real. But it is not a diagnosis. This framework is built around actually figuring out what is wrong.
Ten chapters across four layers
The framework splits AI brand visibility into four layers, each asking a different question. The first three build on each other in order. Layer 4 is the foundation that makes everything else reliable.
The first three layers build on each other. You need to understand your visibility before you can diagnose your perception, and you need both before the influence layer makes sense. Skipping ahead does not save time. It just gives you the wrong answers.
Layer 4 is not a final step. Think of it as the ground everything else stands on. How you design your prompts, how you handle entity data, how you collect responses at scale: all of this shapes whether your numbers from Layers 1, 2 and 3 can actually be trusted.
Say your Share of Voice score is low. You go straight to your Citation Data and nothing looks obviously wrong. You publish more content. The score stays the same. What you missed was the AI Competitive Map, which would have shown you that three brands you had never heard of are dominating the queries you care about. More content was not the fix. Understanding the competitive picture was.
Visibility
Are you in the conversation?
Before you measure how much you appear, you need to know who else is showing up alongside you. The AI’s picture of your market shapes everything, and in most cases it looks quite different from the competitive landscape you think you are in.
AI Competitive Map
The full list of brands the AI connects to your market, built from what it actually says rather than from your existing competitor list. There is almost always a gap between that list and yours. Brands you have never heard of often appear. Brands you consider your main rivals sometimes do not. That gap is where the investigation starts.
AI Share of Voice
How often your brand comes up in AI responses compared to everyone else who gets mentioned. Most tools measure this against a fixed list of competitors you define upfront. Waikay measures it against every brand the AI actually mentions. The difference matters a lot.
Perception
What does AI think you are?
Once you know there is a visibility gap, this layer helps you understand why. It is not about whether you show up. It is about what the AI connects you to when you do. Which topics? How accurately? What kind of brand does it think you are?
This is where most of the diagnostic work happens. Two brands can have the same Share of Voice score and be in completely different positions. One might be associated with five key topics in the market. The other might be strongly tied to just one. And one might be described accurately while the other is full of hallucinations. The number looks the same. The situations are not.
AI Topical Presence
The topics, use cases, and problems the AI connects to your brand, scored on how strong, how broad, and how evenly spread those connections are. This is what explains your Share of Voice score. If your SOV is lower than it should be, Topical Presence shows you exactly what is missing.
Factual Accuracy Rate
How accurately AI describes your brand. The products it attributes to you, the claims it makes about your history and positioning. AI does not hallucinate randomly. It fills in gaps with things that sound plausible, and the less coverage you have, the more likely this becomes. This is a perception problem because it directly shapes how AI presents you to potential buyers.
Influence
Why does AI say what it says?
This is the layer most teams skip, and it is the one that leads to the most expensive mistakes.
Once you know what AI thinks of you, this layer helps you understand why it thinks that. Which sources are shaping its view? How is your content structured in a way that influences the associations it builds? These are the levers you can actually pull to change what the model says.
Entity Map
A map of how your brand’s content is structured and how that structure shapes the associations AI builds. Understanding your entity map tells you why AI connects you to certain topics and not others, and what you need to change to shift those associations.
Citation Data
The sources the AI draws on when it talks about your brand. Which platforms, which types of content, which domains. This tells you where your visibility is actually coming from and, just as importantly, where your competitors are getting cited and you are not. Citations are influence because they directly shape what the model learns to say about you.
Publishing more content fixes a content gap. It does not fix a structural problem with how your content is organised, or change which sources the model is drawing on. Layer 3 tells you which lever to pull before you spend time pulling the wrong one.
Methodology
How do you know your measurements are right?
Layers 1, 2 and 3 cover what to measure. Layer 4 covers how to measure it in a way you can actually rely on.
This is the part most guides leave out because it feels like plumbing rather than strategy. But the plumbing matters. If your prompts are designed badly, your Share of Voice data will be skewed. If your entity analysis is inconsistent, your Topical Presence scores will not reflect reality. If your data collection is noisy, everything downstream will be too.
Layer 4 sits at the end of the guide not because it is an afterthought, but because it is what you come back to when your numbers stop making sense.
Ch. 7 – AI Visibility Channels
Training data versus grounded search. What these two channels are, why they behave differently, and why measuring only one gives you an incomplete picture.
Ch. 8 – Prompt Tracking
How to build prompt sets that give you consistent, unbiased data. The prompts you use shape the results you get, and most teams get this wrong without realising it.
Ch. 9 – NLP and Entity Analysis
How AI groups concepts together, how to map the topics your brand is or is not connected to, and how to spot entity problems at the root rather than the surface.
Ch. 10 – Data Gathering Methods
APIs, scraping, and manual testing. How to collect AI responses at scale without introducing the kind of bias that makes your trend data useless.
Run some prompts. Collect some data. Start building a picture. You do not need a flawless setup from day one. But at some point, usually when your data produces results that do not match what you are seeing in the real world, you will want to come back here and check your foundations.
How to use this guide
You do not have to read it from the beginning
This guide works just as well as a reference as it does a straight read-through. Chapter 0 is the only one worth reading first because it is the map. After that, go to wherever your actual problem is.
I have no idea how AI sees my brand
Start with Chapter 1 (AI Competitive Map). It gives you the clearest starting picture, fastest.
My SOV score is low and I do not know why
Start with Chapter 3 (AI Topical Presence). That is almost always where the answer is.
AI is saying wrong things about my brand
Start with Chapter 4 (Factual Accuracy Rate). Then check Chapter 5 (Entity Map) to understand the structural reasons behind it.
I want to understand the full picture first
Read Chapter 0, Chapter 1 and Chapter 2 in order. That covers the framework, your competitive starting point, and your share of voice baseline.
I need to set up proper measurement infrastructure
Go straight to Layer 4: Chapter 7, Chapter 8, Chapter 9, and Chapter 10. These cover everything you need to build reliable, repeatable data.
Each chapter mentions how Waikay tracks the relevant metric. Those parts are clearly marked and easy to skip if you are building your own setup. The methodology chapters in particular are written to be useful no matter what tools you are using.
