One question in ChatGPT does not tell you how visible your business really is in AI-generated answers.
If the system does not mention your company, it does not necessarily mean that AI never recommends it. And the opposite is also true: one successful recommendation does not mean your business has strong AI visibility.
In one of our practical measurement runs, a company received 0 mentions across 48 checks. Further analysis showed that even a result like this does not yet explain where the actual problem lies.
To get a meaningful picture, you need to:
test different customer queries → compare several AI systems → measure recommendation share → see which competitors AI recommends instead → investigate why
1. Start with Real Customer Queries
The question is not:
“Does ChatGPT know my company?”
A much more useful question is:
“In what share of customer scenarios that matter to my business does AI recommend my company?”
A potential customer might be looking for someone to solve a specific problem, a specialist in their city, a company with a particular area of expertise, a reliable service provider, or someone capable of handling a more difficult job.
These are different decision-making scenarios.
That is why asking three nearly identical questions is not enough:
“What is the best company for X?”
“Recommend a good company for X”
“Which companies do you recommend for X?”
Technically, these are three queries. In practice, they represent almost the same scenario.
The better your query set reflects real situations in which customers choose a provider, the more useful your measurement will be.
2. Check More Than One AI System
For a baseline measurement of a local service business in Poland, we start with four systems:
ChatGPT, Gemini, Perplexity and Google AI
ChatGPT is difficult to ignore because of the scale of its audience. According to OpenAI, more than 900 million people use ChatGPT every week in 2026.
But checking only the market leader is not enough.
Gemini is Google's separate AI system. Perplexity operates as an independent answer and search system and can also be useful when investigating sources. Google AI should be checked separately because users can receive AI-generated answers directly within Google Search.
The key point is that the same business can have very different visibility across different AI systems.
A result from ChatGPT therefore cannot automatically be treated as a result for AI search as a whole.
These four systems are not a universal rule for every business. The scope of measurement should reflect the market and audience of the company being analysed.
3. Measure More Than Mentions
Suppose a company appears in three out of ten queries.
That gives us 30%.
It is a useful starting metric, but the number alone explains very little.
You also need to examine:
- which queries trigger the company
- whether AI actually recommends it or merely mentions it
- which businesses are recommended instead
- which competitors appear repeatedly
- how AI explains its choices
- whether results differ between systems
- which sources and external signals can be identified
Instead of just a visibility percentage, you begin to see how AI positions the company within its competitive landscape.
4. Measurement Is Where Diagnosis Begins
Collecting AI answers and putting them into a spreadsheet is not enough.
If AI repeatedly recommends one particular competitor — why?
If it does not mention your company — what might be missing?
If one system sees the company and another does not — what could explain the difference?
What sources and supporting evidence sit behind the businesses that AI recommends more often?
At this point, measurement stops being a mechanical count of mentions and becomes diagnostic work based on analysing the answers.
5. Why Even 0 Out of 48 Is Not Yet a Diagnosis
In one of our practical measurement runs, we tested 12 customer queries across four AI systems — 48 individual answers in total.
The company did not appear in any of them.
The first conclusion seemed obvious:
0 out of 48 — AI does not recommend the company
But we went further.
When one of the systems was given the company's name and website directly, it was able to understand the business correctly, including its specialisation, offer and strengths.
That changed the diagnostic question:
Why can AI understand the company when we show it the website directly, but fail to surface it independently when a potential customer is looking for that type of provider?
In this case, the measurement led us to investigate separately how the relatively new website was being discovered by search engines and how it was represented within Google's infrastructure.
So 0 out of 48 was a measurement result, not an explanation of the cause.
Two companies can have exactly the same symptom — “AI does not recommend us” — for completely different reasons.
6. Do Not Give AI Your Company Name in the Initial Measurement
There is a fundamental difference between asking:
“Which companies do you recommend for X?”
and:
“What do you think about company Y?”
In the second question, you have already given the system the company's name.
That question is useful too. It can help determine whether AI understands what the business does and what information it can find about it.
But that is a different layer of diagnosis.
It does not answer the key question of the initial measurement:
Will AI surface your company on its own when a potential customer is looking for a provider?
That is why the initial measurement should use queries that do not contain the brand name.
What Should an AI Visibility Measurement Actually Show?
A useful measurement should answer more than “How many times did AI mention us?”
It should show:
where AI recommends the company → where it chooses competitors → who those competitors are → how the systems explain their choices → what needs to be investigated next
Only then does it make sense to decide what should actually be changed on the website, across external platforms and in the information about the business available online.
That is why at KAN we do not start a collaboration by promising to “get your company promoted in ChatGPT.”
First, we establish the baseline: we test real customer scenarios across ChatGPT, Gemini, Perplexity and Google AI, then analyse which businesses the systems recommend and why.
Only after that diagnosis do we determine what may actually be preventing a particular business from appearing more often in AI-generated answers.
Want to check whether AI recommends your business?
We begin by measuring real customer scenarios and diagnosing what affects the company’s visibility.
BOOK A CALL