AI Visibility Testing was created by me, John Greer, a UK website owner with years of practical experience building WordPress websites, publishing content and trying to understand how people discover information online.
My interest in AI visibility developed while I was conducting practical Google AI Mode experiments.
I initially wanted to understand how Google AI Mode found, interpreted and cited information from websites. That led me to ask realistic questions, examine the resulting answers and follow the citations back to their sources.
As the experiments continued, I noticed something particularly important: asking the same question more than once did not necessarily produce the same answer.
The businesses mentioned could change. Different websites could be cited. Information included in one response could disappear from the next. Sometimes the overall conclusion remained similar even though the supporting sources changed substantially.
That raised the question at the heart of this website:
If one AI search can produce a different result from the next, how can a business reliably understand where and when its brand appears?
From Individual Searches to Systematic Testing
Finding your website mentioned or cited in an AI-generated answer can be encouraging. Failing to find it can be concerning.
Neither result means very much when viewed on its own.
A single appearance does not establish consistent visibility. Equally, being absent from one response does not prove that a brand will always be absent.
A more meaningful picture requires repeated testing.
That means recording the exact prompt, running it several times, separating brand mentions from website citations and comparing which competitors, sources and recommendations appear.
AI Visibility Testing was created to explore and explain that process.
What This Site Investigates
The experiments and guides published here examine subjects such as:
- How an AI-generated response is assembled
- How claims and citations relate to different sections of an answer
- Whether cited sources genuinely support the statements made
- What changes when an identical prompt is repeated
- How results differ between Google AI Mode and ChatGPT Search
- How detailed, scenario-based prompts affect the information required
- How query fan-out may divide a complicated question into connected parts
- How to distinguish a brand mention from a brand-owned website citation
- How to calculate a clearly defined measure of AI share of voice
- When manual testing becomes too cumbersome to manage consistently
The purpose is not to declare which AI platform is best. It is to understand what each test reveals and how much confidence should reasonably be placed in the result.
An Evidence-Led Approach
I am not an AI scientist, and I do not claim to possess a guaranteed method for making a website appear in AI-generated answers.
My experience comes from building websites, examining search performance and conducting practical experiments with real AI search responses.
The site follows several basic principles:
- Exact prompts should be recorded.
- Tests should be repeated.
- Results should not be discarded merely because nothing interesting happened.
- Brand mentions and website citations should be counted separately.
- Citations should be checked against the claims they appear to support.
- Observations should be distinguished from interpretations.
- Theories should not be presented as proven facts.
- Limitations should be acknowledged.
Where an experiment suggests a possible explanation, it will be described as a possibility—not a guaranteed cause or optimisation technique.
Each experiment aims to separate three things:
What was observed
The results that actually appeared during the recorded tests.
What those results may suggest
A reasonable interpretation that could justify further investigation.
What the experiment cannot prove
The conclusions that would go beyond the available evidence.
Manual Testing Before Automation
A small AI visibility test does not require specialist software.
A website owner can select a handful of realistic customer questions, run them several times and record the brands and citations that appear in a spreadsheet.
That manual process is useful because it makes the measurement understandable. You can see exactly what has been counted and why one result may differ from another.
However, the workload grows quickly.
Monitoring more prompts, competitors and platforms—and repeating the exercise regularly—can create hundreds of individual responses to collect and compare.
This site therefore also examines the point at which automated monitoring software may become worthwhile.
Automation is not presented as a way to guarantee greater visibility. Its practical value is its ability to perform and organise repeated checks that would otherwise take too much time to complete manually.
Commercial Transparency
AI Visibility Testing may recommend relevant monitoring services or other tools where they offer a practical way to perform work explained on this site.
Some recommendations may use affiliate links, which means I could receive a commission if someone purchases through them. Where this applies, it will be disclosed clearly.
An affiliate relationship will not turn a theory into a fact or make a tool appropriate for everyone. Manual methods will be explained so that readers can understand what a service does before deciding whether they need it.
The Aim of AI Visibility Testing
The aim is not to promise that a website can be made to appear in a particular AI-generated response.
It is to help website owners, businesses, marketing teams and agencies ask better measurement questions:
- Where does the brand appear?
- How often does it appear?
- Is it mentioned, recommended or cited?
- Which competitors appear alongside it?
- Which sources support the answer?
- What changes when the test is repeated?
- How much testing is needed before a pattern becomes meaningful?
AI visibility is not a fixed ranking position. It is a changing pattern that must be observed carefully.
AI Visibility Testing exists to make that process clearer, more systematic and easier to understand.
Contact
If you have a question about the site, its methodology or a published experiment, you can contact me at:
About AI Visibility Testing
AI Visibility Testing was created by me, John Greer, a UK website owner with years of practical experience building WordPress websites, publishing content and trying to understand how people discover information online.
My interest in AI visibility developed while I was conducting practical Google AI Mode experiments.
I initially wanted to understand how Google AI Mode found, interpreted and cited information from websites. That led me to ask realistic questions, examine the resulting answers and follow the citations back to their sources.
As the experiments continued, I noticed something particularly important: asking the same question more than once did not necessarily produce the same answer.
The businesses mentioned could change. Different websites could be cited. Information included in one response could disappear from the next. Sometimes the overall conclusion remained similar even though the supporting sources changed substantially.
That raised the question at the heart of this website:
If one AI search can produce a different result from the next, how can a business reliably understand where and when its brand appears?
From Individual Searches to Systematic Testing
Finding your website mentioned or cited in an AI-generated answer can be encouraging. Failing to find it can be concerning.
Neither result means very much when viewed on its own.
A single appearance does not establish consistent visibility. Equally, being absent from one response does not prove that a brand will always be absent.
A more meaningful picture requires repeated testing.
That means recording the exact prompt, running it several times, separating brand mentions from website citations and comparing which competitors, sources and recommendations appear.
AI Visibility Testing was created to explore and explain that process.
What This Site Investigates
The experiments and guides published here examine subjects such as:
- How an AI-generated response is assembled
- How claims and citations relate to different sections of an answer
- Whether cited sources genuinely support the statements made
- What changes when an identical prompt is repeated
- How results differ between Google AI Mode and ChatGPT Search
- How detailed, scenario-based prompts affect the information required
- How query fan-out may divide a complicated question into connected parts
- How to distinguish a brand mention from a brand-owned website citation
- How to calculate a clearly defined measure of AI share of voice
- When manual testing becomes too cumbersome to manage consistently
The purpose is not to declare which AI platform is best. It is to understand what each test reveals and how much confidence should reasonably be placed in the result.
An Evidence-Led Approach
I am not an AI scientist, and I do not claim to possess a guaranteed method for making a website appear in AI-generated answers.
My experience comes from building websites, examining search performance and conducting practical experiments with real AI search responses.
The site follows several basic principles:
- Exact prompts should be recorded.
- Tests should be repeated.
- Results should not be discarded merely because nothing interesting happened.
- Brand mentions and website citations should be counted separately.
- Citations should be checked against the claims they appear to support.
- Observations should be distinguished from interpretations.
- Theories should not be presented as proven facts.
- Limitations should be acknowledged.
Where an experiment suggests a possible explanation, it will be described as a possibility—not a guaranteed cause or optimisation technique.
Each experiment aims to separate three things:
What was observed
The results that actually appeared during the recorded tests.
What those results may suggest
A reasonable interpretation that could justify further investigation.
What the experiment cannot prove
The conclusions that would go beyond the available evidence.
Manual Testing Before Automation
A small AI visibility test does not require specialist software.
A website owner can select a handful of realistic customer questions, run them several times and record the brands and citations that appear in a spreadsheet.
That manual process is useful because it makes the measurement understandable. You can see exactly what has been counted and why one result may differ from another.
However, the workload grows quickly.
Monitoring more prompts, competitors and platforms—and repeating the exercise regularly—can create hundreds of individual responses to collect and compare.
This site therefore also examines the point at which automated monitoring software may become worthwhile.
Automation is not presented as a way to guarantee greater visibility. Its practical value is its ability to perform and organise repeated checks that would otherwise take too much time to complete manually.
Commercial Transparency
AI Visibility Testing may recommend relevant monitoring services or other tools where they offer a practical way to perform work explained on this site.
Some recommendations may use affiliate links, which means I could receive a commission if someone purchases through them. Where this applies, it will be disclosed clearly.
An affiliate relationship will not turn a theory into a fact or make a tool appropriate for everyone. Manual methods will be explained so that readers can understand what a service does before deciding whether they need it.
The Aim of AI Visibility Testing
The aim is not to promise that a website can be made to appear in a particular AI-generated response.
It is to help website owners, businesses, marketing teams and agencies ask better measurement questions:
- Where does the brand appear?
- How often does it appear?
- Is it mentioned, recommended or cited?
- Which competitors appear alongside it?
- Which sources support the answer?
- What changes when the test is repeated?
- How much testing is needed before a pattern becomes meaningful?
AI visibility is not a fixed ranking position. It is a changing pattern that must be observed carefully.
AI Visibility Testing exists to make that process clearer, more systematic and easier to understand.
Contact
If you have a question about the site, its methodology or a published experiment, you can contact me at: