The New Visibility Metrics in the Age of AI

The shift: AI has moved to the top of the funnel

Buyers increasingly ask AI platforms complex questions and receive structured answers that include:

Instead of opening ten browser tabs and stitching together understanding themselves, they receive a synthesised answer in one response.

This is not a small change; it’s a structural shift in how information is discovered, evaluated and prioritised.

And it creates a new measurement problem.

Why AI discovery breaks traditional measurement

Traditional search measurement assumed a stable structure:

AI discovery though behaves very differently:

a. Visibility often happens inside the answer

In traditional search, visibility meant appearing in the results page.

In AI discovery, visibility often means:

Sometimes the user clicks — often they don’t.

Traffic becomes a lagging indicator of influence, not the primary one.

b. Click behaviour drops when AI answers appear

Early research confirms this pattern; when AI-generated summaries appear in search results:

In other words people read the answer, absorb the insight, and then move on.

Your brand may influence the buyer’s thinking — without your analytics ever registering a visit.

c. AI answers are probabilistic

Search rankings are relatively stable. AI answers are not.

Responses can change depending on:

Ask the same question several times and you may receive slightly different answers. Which means optimisation becomes less about achieving a single ranking position and more about increasing the probability of inclusion.

d. Answers are increasingly personalised

Some AI systems now retain context and user memory. This means two people asking the same question may receive different answers because the system factors in:

There isn’t one fixed answer anymore — there are many different versions.

Which reinforces a key principle: AI visibility is a probability problem, not a ranking problem.

e. Agentic AI adds an infrastructure layer

The next shift goes beyond AI chat answers: “AI agents” are AI systems that can carry out tasks on their own — such as evaluating vendors.

In this environment, visibility is no longer just about appearing inside AI answers. It is also about ensuring AI systems can access, understand and use your expertise when they are gathering information on their own.

This is where the idea of an infrastructure layer comes in.

Traditional AI discovery (for example inside ChatGPT or Perplexity) usually works like this:

Optimisation focuses on being included in that answer.

Agentic AI though changes the process. Instead of simply answering a question, an AI agent can be given a goal — such as researching vendors, comparing tools, or gathering evidence for a decision. To do this it may:

In other words, the AI is not just producing an answer; it is actively navigating the web as part of a workflow.

That is why this introduces an infrastructure layer — your expertise must not only be summarisable by an AI model, it must also be easy for AI systems to retrieve, interpret and work with when they are performing tasks.

This is different from traditional GEO optimisation for AI chat interfaces, which mainly focuses on helping models:

Agentic discovery requires something broader — content must also be:

This is where Discovery Engineering becomes important. It means deliberately designing content and information structures so AI systems — whether answering questions or performing research tasks — can confidently recognise and use your expertise.

The measurement challenge

All of this creates a new problem for marketing teams. Traditional dashboards measure:

But those metrics capture only the final stage of discovery.

They do not capture whether a brand influenced the answer that shaped the buyer’s thinking.

This is why many teams are experiencing a strange contradiction: Traffic may fall as AI answers reduce clicks, but the visitors who do arrive often come with stronger intent.

This is where a structured AI Visibility Measurement Stack becomes essential.

Core Framework: The AI Visibility Measurement Stack

To measure AI discovery properly, we need a framework that reflects how AI retrieval actually works.

The AI Visibility Measurement Stack provides a way to measure influence across four connected layers.

Each layer measures a different dimension of AI visibility — from basic retrieval to real commercial impact — mapping directly to the operational pillars of AI visibility strategy.

The primary sources for measuring AI visibility

Because AI discovery is still evolving, measurement currently relies on a mix of approaches.

These include:

Prompt testing

Systematically testing prompts across multiple AI systems to observe retrieval behaviour.
This provides insight into:

AI visibility monitoring tools

A new generation of tools is emerging to track AI citations and mentions. Examples include:

These platforms attempt to quantify how often a brand appears inside AI-generated answers.

Topic coverage analysis

Another diagnostic approach focuses on content coverage rather than citations. This examines whether a brand’s content answers the key questions AI systems rely on when explaining a topic.

In practice, it involves mapping common industry questions and checking whether your site provides clear content addressing them.

While this does not measure visibility directly, it helps explain why a brand may or may not appear in AI-generated answers, highlighting gaps that reduce the likelihood of retrieval.

Visibility metrics tell you what AI is retrieving.
Topic coverage analysis helps explain why.

Entity and knowledge graph signals

AI systems try to understand who a brand is and what it is known for, not just the words on a page. In technical terms, this is called an “entity” — a clearly identifiable organisation, person, or concept.

These signals can be monitored by checking whether your brand appears consistently across sources that AI systems commonly reference, such as company directories, knowledge bases, industry sites and news articles. They can also be assessed by reviewing your own website to ensure it includes structured data (called schema) that clearly identifies your organisation and its topics.

Typical signals to track include:

Together, these signals indicate whether AI systems can confidently recognise your brand as a trusted organisation associated with a specific topic.

Avoiding the new vanity metrics

Every marketing shift produces new vanity metrics and AI discovery already has several.

Examples include:

These signals may look impressive but often lack strategic meaning.

The real goal is not a single appearance; it is consistent retrieval across prompts, topics, and engines.

In other words: not “Did we appear once?”, but: “Are we reliably included when the market asks questions we should own?”

The real question: Are you visible when AI explains your market?

AI-mediated discovery is changing how influence works at the top of the funnel.

For years, marketing teams focused on a simple question: Where do we rank?

But rankings were only ever a proxy for something more important — visibility during the research phase.

AI has simply made that phase more explicit.

Instead of scanning pages and assembling their own understanding, buyers now receive structured answers that already contain:

Which means the brands included in those answers shape how the market understands a problem — and which solutions are considered credible.

The ones that are absent don’t just lose traffic. They lose influence.

This is why measuring AI discovery properly matters.

Traditional analytics tells you what happened after someone arrived. The AI Visibility Measurement Stack helps you understand something earlier and far more important:

In other words, it measures influence before interaction.

For content-led B2B organisations whose growth depends on expertise being discovered and trusted, that distinction is becoming critical.

Because as discovery shifts from pages to answers, one thing becomes increasingly clear:
Inclusion is the new visibility signal.

And the organisations that learn to measure and engineer that inclusion early will gain a significant advantage — not just in traffic, but in how their markets form opinions and make decisions.