"Recommend someone to build a website in Ostrava." "What accounting software is good for a small business?" "Where can I buy good Moravian wine?" People increasingly ask questions like these of AI assistants instead of Google. If an answer omits you, you are absent from that particular recommendation. This alone does not prove a lost customer or a problem with your whole website. In this article I show a simple way to find out whether ChatGPT recommends you (and other AI assistants too), how to evaluate the results and what to do if they don't know you.
Choose 10 to 20 real customer questions without your business name. Record the product, model or mode, date, answer and cited sources for each question. Repeat the questions several times. This produces a small repeatable sample, not a score of your overall visibility or proof of lost sales.
Why care what AI says about you?
An AI answer may present a limited selection of businesses. Names and cited sources vary with the question, product and search mode. Treat your records as a small visibility sample, not a survey of all customers. Compare AI traffic volume and lead quality in your own analytics and CRM; one study cannot establish the result for every business. I've collected the numbers and sources in the article AI search is already sending customers.
How do you find out whether AI assistants recommend you?
1. List the questions your customers ask
The key is that the questions don't include your business name. If you ask "what is company XY", AI will find you easily. What you want to know is whether it mentions you when the customer doesn't yet know you exist.
Good questions fall into three types:
| Question type | Example |
|---|---|
| Recommendation | "Who builds custom e-shops in Ostrava?" |
| Comparison and choice | "Shoptet or a custom e-shop for a small business?" |
| Price and decision | "How much does a business website cost and who should build it?" |
Ten to twenty questions are a starting point for a small test, not a representative survey. Take them from what customers ask you on the phone and by email.
2. Ask the questions in several tools
Each AI assistant works with different sources, so test more than one:
- ChatGPT, ideally in a temporary chat with personalization turned off, recording the actual setting,
- Perplexity, which shows its sources clearly next to each answer,
- Google, where you watch the AI Overview above the results,
- optionally Microsoft Copilot and Gemini.
Watch out for personal bias: if you often use ChatGPT to work on your own business, it may "know" more about it than it would for a stranger. A temporary chat alone is not sufficient: current documentation allows it to be personalized. Select the unpersonalized option before the first question and record it.
3. Repeat each question
AI assistants may give different answers when you repeat a question. A single answer is just one observation. Repeat each question three times and note how many times you were mentioned. "2 out of 3" is more useful than "yes, once".
4. Record the results in a table
A simple table is enough:
| One query record | Fictional result |
|---|---|
| Question | Who builds custom e-shops in Ostrava? |
| Product, model and mode | ChatGPT; record the exact model and search setting |
| Date and location | Record date, language and location |
| Own business mentioned | No; 0 out of 3 repetitions |
| Other businesses | Company A, Company B (fictional example) |
| Source | Save the exact cited URL and answer |
The records of other mentioned businesses and cited sources are particularly useful. They show you who AI recommends instead of you and where it gets that from. If it repeatedly cites a directory, comparison site or article, check its relevance and the accuracy of its information. A listing there does not entitle you to a recommendation.
5. Repeat the test once a month
Same questions, same tools, same process. A comparison over time shows how this sample changes. It does not establish the cause: the model, available sources and product may also change.
Worked example: One query, a repeatable record
worksheet
| Field | Run 1 | Run 2 |
|---|---|---|
| Date + time | To record | To record |
| Product + model + mode | To record | To record |
| Business mention | Not checked | Not checked |
| Cited source + URL | Not checked | Not checked |
Blank protocol; contains no actual AI recommendation.
What affects who an AI assistant recommends?
There is no published universal recipe for business recommendations across AI products. The following four areas are practical checks of information availability and clarity, not a proven ranking of factors or a guarantee of citation.
- Search visibility. AI assistants with search look information up on the web. ChatGPT search also works with Bing results, and Google AI Overviews draw on Google's index. A website that isn't visible in these search engines is hard for AI to find.
- Content that can be quoted. Clearly described services, prices, industry and service area. If your website doesn't say clearly what you do and for whom, AI won't say it for you. That is how I plan websites that state the offer plainly, with services, prices and service area in words. I explain how to write this kind of content in the article how to write content for AI search.
- Mentions on other websites. Reviews, directories, industry articles, interviews, lists of recommended businesses. Check whether cited independent sources contain correct information about your business. Their presence alone does not guarantee a recommendation.
- Consistent details. The same name, address, industry and offer on your website, your Google profile, Firmy.cz and other directories. Conflicting details can confuse customers and systems that reuse the information; correcting them does not yield a measurable probability of a mention.
What should you do if AI assistants don't know you?
Start with the simplest steps:
- Make sure search engines can find you. Your site in Google Search Console and Bing Webmaster Tools, with the important pages indexed.
- Add clear business information to your website. What exactly you do, for whom, where, for how much and who is behind the business. An "about me" or "about us" page with concrete facts is a valuable source for AI.
- Align the details in your profiles. Google, Firmy.cz, industry directories, social media.
- Get mentioned where AI looks. Check the "source cited" column in your table. If AI repeatedly cites a particular directory or comparison, get listed there if it makes sense.
- Collect reviews. Regularly and honestly, ideally with a concrete description of what you did for the customer.
- Write content for the questions from your test. If customers ask "how much does X cost and who does it" and your website has no answer, AI has nothing to cite.
You'll find a more detailed technical process (Bing, the llms.txt file, structured data) in the article how to get your business into AI search.
Worked example: A business needs a specific description
ORIGINAL EXAMPLE · FICTIONAL DATA
Before
- "An innovative team with a personal approach."
- Service, location and contact are unclear.
After
- "North Workshop repairs bicycles in Ostrava. Request brake and gear servicing through the contact form."
- Verify against the actual offer and contacts.
Fictional business. Clearer copy does not guarantee AI recommendations.
What to avoid
- Testing only your own business name. It shows whether AI knows you, not whether it recommends you.
- Drawing conclusions from one answer. Answers change; you need a repeated test.
- Buying "AI mentions" or listing your business in dozens of dubious directories. The quality of a source matters more than the number.
- Writing text for AI instead of people. Write clearly and verify facts; clear wording still does not guarantee correct model interpretation.
Frequently asked questions
How do I find out whether ChatGPT recommends me?
In an unpersonalized temporary chat, ask it 10 to 20 questions your customers ask, without your business name. Repeat each question three times and note how many times it mentioned you and who it recommended instead. Run the same test in Perplexity and Google.
Why does ChatGPT mention me once and not the next time?
AI assistants don't give the same answer every time. It depends on the exact wording of the question, the sources found at that moment and the randomness in how the model composes its answer. That's why it makes sense to track the share of repetitions in which you're mentioned, not a single answer.
Can you buy a mention in ChatGPT?
A paid advertisement or listing is not the same as an independent recommendation in an answer. Ask the supplier to identify the placement and paid disclosure. These steps cannot guarantee an organic recommendation across AI assistants.
How often should I test my AI visibility?
Monthly repetition is a practical starting point, not a validated standard. After an important offer or product change, run an additional comparable test and record the changed conditions.
How long does it take before AI starts recommending me?
There is no reliable deadline. Source availability, indexing and answer generation change, and even a good website does not guarantee a recommendation. Track repeated observations and actual visits or enquiries.
Want to know where you stand?
If you'd like me to review your business's visibility in Google and AI assistants and suggest what to change, take a look at SEO and AI search preparation. Preparation for AI search is part of it. For a quick first overview, you can try the free website check.
Sources and links
Sources and limits of the test
Google requires an indexed page eligible for Search snippets for inclusion as a supporting link in AI features. Inclusion is not guaranteed. OpenAI describes OAI-SearchBot access for content discovery. These requirements do not establish that three repetitions are sufficient; the repetition count here is a practical small-test proposal.