AI Visibility Testing: How to Test Your Business in AI Search

AI visibility testing is the process of checking how a business, brand, product or website appears in AI-generated answers.

That can include Google AI Mode, ChatGPT and other AI search platforms.

At first, the obvious question seems to be:

Does my business appear?

But a single AI answer can be misleading.

A business may be recommended in one response and absent from the next. Its website may be cited without the business itself being recommended. A competitor may dominate a broad query but disappear when the customer describes a more specific requirement.

That means useful AI visibility testing needs to go further than checking one prompt once.

The aim is to understand:

which customer searches are actually worth being visible for;
how consistently your business appears;
whether it is mentioned, cited or genuinely recommended;
how competitors perform across the same searches;
how accurately AI describes your business;
and how those patterns change across prompts, platforms and time.

This page brings together the practical guides and experiments on AI Visibility Testing so you can follow the process from the beginning.

Start Here: A Practical AI Visibility Testing Process

You do not need specialist software to run your first AI visibility tests.

Start with a small number of important customer situations, test them carefully and record what happens.

1. Choose Searches That Actually Matter

The quality of an AI visibility test depends heavily on the prompts being measured.

A business could appear frequently across broad informational searches while being almost invisible when a potential customer describes the particular problem that business is well placed to solve.

That is why prompt selection should begin with the customer rather than simply creating the longest possible list of keywords or questions.

Start with:

How to Choose AI Search Prompts That Actually Matter to Your Business

Then, if you want to build a more structured set of prompts:

How to Build Your First AI Visibility Prompt List From the Customer Journey

and:

How to Build an AI Visibility Query Set Around Real Customer Situations

The objective is not to predict every possible question someone could ask.

It is to build reasonable coverage of the customer needs and situations that matter to the business.

2. Repeat Important Searches

Once you have chosen a worthwhile prompt, do not assume that one response represents your normal AI visibility.

AI-generated answers can change even when the wording of the query stays exactly the same.

Run the same important searches repeatedly under broadly similar conditions and record which businesses appear.

Our practical starting guide is:

How to Run a Repeatable Manual AI Visibility Test

A related question is how many repetitions you actually need.

There is no universal number that automatically makes an AI visibility test statistically reliable. The appropriate level of repetition depends on what you are trying to learn.

For a practical manual check, the objective may simply be to see whether an apparently strong recommendation is stable, occasional or absent.

See:

How Many Times Should You Repeat an AI Search When Testing Brand Visibility?

The important distinction is that one AI answer is an observation, not a measurement of consistent visibility.

3. Separate Mentions, Citations and Recommendations

Not every appearance inside an AI answer means the same thing.

An AI system might:

recommend your business as a suitable choice;
mention the brand somewhere in its explanation;
use your website as a supporting source;
cite a page simply to verify one fact;
or describe a competitor while linking to your website for background information.

Recording all of those outcomes as simply “visible” can hide important differences.

Read:

Mentioned, Cited or Recommended: What Kind of AI Visibility Do You Have?

We also tested the relationship directly:

I Compared AI Citations With Recommendations Across 10 Google AI Mode Searches

A cited website is not automatically a recommended business.

That distinction matters when deciding what your visibility actually means.

4. Measure How Consistently You Appear

Once you have repeated a prompt several times, you can begin turning those observations into useful measurements.

For example, if your business is recommended in three out of five identical searches, that tells you something different from appearing once or five times.

A simple way to start is:

How to Calculate AI Recommendation Share of Voice Manually

But one prompt only measures one customer situation.

To understand broader visibility, you also need to see what happens when the customer’s requirements change.

Our worked example does exactly that:

How to Measure AI Share of Voice Manually: What 25 Searches Revealed

This is an important distinction.

Repetition measures consistency within a query.

Testing different customer situations measures coverage across the market you care about.

A useful AI visibility programme will normally need both.

5. Look at Competitors, Not Just Your Own Score

A visibility percentage needs context.

Suppose your business appears in 30% of the AI answers you test.

Is that good?

It depends on what happens to everybody else.

If the leading competitors appear 80% of the time, your 30% may indicate a substantial gap.

If most competitors appear only occasionally, 30% may look much stronger.

Competitor analysis can also reveal where another business is winning.

You may find that a competitor rarely appears for broad searches but consistently appears when customers mention one particular requirement.

That can give you a useful area to investigate:

What does that competitor specialise in?
What evidence appears on its website?
Which customer questions does it answer particularly well?
Is AI associating that business with a need that it does not associate with yours?

AI visibility testing should therefore be used to investigate patterns rather than simply to produce a headline score.

6. Check What AI Actually Says About Your Business

Appearing is not necessarily enough.

An AI system might recommend your business while:

describing the wrong services;
misunderstanding who you serve;
overlooking an important specialism;
inventing a feature;
confusing your business with another company;
or presenting outdated information.

That creates a different kind of visibility problem.

A useful sense check is:

How to Test If AI Understands Your Brand: A Simple Sense Check

The objective is not merely to ask:

Did we appear?

It is also to ask:

Did the AI understand why we might be relevant to this customer, and did it represent us accurately?

The Experiments Behind the Method

The practical guidance on this site has developed from running real AI searches and comparing what happened.

The purpose of those experiments is not to claim that every AI platform, query or industry will behave identically.

Instead, the process is:

Test something → record what happened → identify a pattern → test the pattern again → turn the useful parts into practical guidance.

Here are some of the experiments that have shaped the approach.

The Same Google AI Mode Search Produced Different Recommendations

One of the earliest experiments repeated exactly the same detailed product search ten times.

Six different products appeared across the ten responses. Four different products occupied first place, while only one appeared in every run.

Read the experiment:

I Ran the Same Google AI Mode Product Search 10 Times—Here’s How the Recommendations Changed

This was one of the clearest demonstrations of why a one-off AI visibility check can be misleading.

Google AI Mode and ChatGPT Did Not Recommend the Same Businesses

Repeating a search on one platform is only part of the picture.

We gave Google AI Mode and ChatGPT the same detailed buying requirement five times each.

The recommendation patterns were different.

One accountancy firm ranked first in all five Google AI Mode searches but never ranked first in ChatGPT. Another ranked first four times in ChatGPT but never first in Google AI Mode.

Read:

Google AI Mode vs ChatGPT: I Asked the Same Buying Question Five Times

So instead of simply asking:

Are we visible in AI search?

it may be more useful to ask:

Where are we visible, for which customer situations, and how consistently?

Citations and Recommendations Are Not the Same Thing

Another experiment examined which websites Google AI Mode cited and compared those sources with the businesses actually being recommended.

The two groups were not identical.

A website could help support an AI-generated answer without its business being one of the recommended choices.

That is why citation visibility, brand visibility and recommendation visibility should not automatically be combined into one measure.

Read the 10-search citation and recommendation experiment

Test Customer Situations, Not Just Keywords

One of the strongest themes to emerge from the testing is the importance of the customer’s actual situation.

Compare:

best accountant Birmingham

with:

I run a 25-person manufacturing company in Birmingham, use Xero and need year-end accounts, corporation tax, payroll and monthly management accounts. I would prefer fixed monthly fees. Which firms should I consider?

Both concern Birmingham accountants.

But the second search gives the AI considerably more information about what would make a provider suitable.

Changing meaningful customer requirements can change the businesses an AI system recommends.

That is why a useful prompt portfolio should explore the customer journey and the circumstances that genuinely alter what a good answer looks like.

You can explore this area further in the Customer Journeys section of the site.

Commercial Value Matters Too

A business could achieve a high AI visibility percentage across searches that have relatively little commercial importance.

Another business might appear less frequently overall but perform extremely well when potential customers describe precisely the problem it specialises in solving.

Those results should not automatically be valued equally.

That is the thinking behind the:

RAMP Framework: Measuring the Commercial Value of AI Visibility

RAMP considers four questions:

R — Relevance: Was this a customer search worth appearing for?

A — Appearance: Did the business actually appear, and how consistently?

M — Mention Quality: How was the business presented?

P — Pathway: Was there a practical route for the customer to move towards the business?

AI visibility is therefore not simply about accumulating mentions.

The more useful question is whether you appear for the right searches, in the right way, with a meaningful opportunity for that visibility to matter.

AI Visibility Testing Is an Ongoing Process

There is unlikely to be one perfect collection of prompts that you create once and monitor forever.

Testing itself teaches you things.

You may discover:

queries that seemed promising but turn out to have little commercial value;
customer situations you had not originally considered;
new competitors;
recurring AI misunderstandings;
topics where your business appears particularly strongly;
areas where competitors repeatedly outperform you;
or differences between AI platforms.

Your testing should therefore evolve as your understanding improves.

Keep important core prompts so that you can observe changes over time, but continue exploring and refining the wider set.

AI visibility testing is better thought of as a process than a score.

When Does AI Visibility Software Become Useful?

Manual testing is extremely useful at the beginning.

It forces you to look closely at individual answers and understand what you are actually measuring.

But the workload grows quickly.

Imagine monitoring:

20 customer prompts;
across four AI platforms;
against several competitors;
repeatedly;
over several months.

Manual testing soon turns into hundreds or thousands of observations.

That is where automated AI visibility monitoring starts to make practical sense.

Software does not remove the need to choose good prompts or interpret the results.

It makes worthwhile testing easier to repeat at scale.

You can explore the platforms I have tested in the AI Visibility Tools section.

What AI Visibility Testing Cannot Tell You

There are important limits to this work.

These experiments cannot reveal the complete internal processes used by Google, OpenAI or other AI systems.

They do not establish that one website change caused a particular AI recommendation.

And an experiment involving one query, market or platform should not automatically be turned into a universal ranking rule.

AI systems change.

Markets differ.

Competitors differ.

Prompts differ.

The purpose of AI visibility testing is therefore not to promise:

Do this and you will rank in AI.

It is to replace assumptions with observations.

What happened?

Does it happen repeatedly?

When does it change?

Which competitors appear?

How are we being represented?

Which customer situations seem to matter most?

Those are questions businesses can actually investigate.

A Simple Place to Begin

If you want to start testing your own business, you do not need to make it complicated.

Begin with this sequence:

Then expand the test gradually as you learn more about your customers, competitors and AI visibility.

The objective is not to monitor everything.

It is to understand your visibility where it actually matters.

Explore More AI Visibility Testing

If you want to go deeper, you can explore the other main areas of the site:

Customer Journeys — learn how to map customer problems, questions and buying decisions into useful AI visibility prompts.

RAMP Framework — assess the commercial value of your AI visibility rather than simply counting appearances.

AI Visibility Tools — explore software for monitoring prompts, mentions, citations, recommendations and competitors when manual testing becomes impractical.

Or browse the full AI Visibility Testing archive for all practical guides, experiments and ongoing research.

The central principle is simple: test the customer situations that matter, repeat the important searches, record what actually happens, and allow the evidence to improve the way you test.