Short answer: Write 8 to 10 buyer questions that do not name your business (“best X for Y”, “alternatives to Z”, “how much does X cost?”), ask each in a fresh chat on two or three assistants, and record whether you are named, in what position, who is named instead and whether what it says about you is correct. Repeat weekly, because answers vary and one run proves little. Unvibe automates this: it asks your buyer questions, records your position and who is named instead, and repeats it every week.
Why you cannot just ask “do you know my business?”
Asking an assistant about your business by name tests what it remembers, not whether it recommends you. A buyer who does not know you yet asks the question you do not appear in: “What’s the best invoicing app for freelancers?” That is the question to test, and the answer is the one that wins or loses the customer.
The method
1. Write 8 to 10 buyer questions
Think of what someone types before they know you exist. Cover four types:
| Type | Example (for a fictional invoicing app) |
|---|---|
| Best-of | What is the best invoicing app for freelance designers? |
| For a specific person | Which invoicing software is good for a freelancer with three clients? |
| Alternatives | What are some alternatives to a spreadsheet for sending invoices? |
| Cost and features | How much does invoicing software for freelancers cost, and is there a free plan? |
Do not name your business in any of them. If you can, use the exact wording customers use in emails and calls.
2. Ask each one in a fresh chat
Open a new conversation for every question, so the earlier answers do not shape the next. Try two or three assistants, because they behave differently. If you can, also ask the same question twice: answers vary from run to run.
3. Record the result in a sheet
One row per question and assistant. These columns are enough:
| Question | Assistant | Named? | Position | Named instead | Facts right? |
|---|---|---|---|---|---|
| Best invoicing app for freelance designers? | Assistant A | No | – | Three competitors | – |
| Best invoicing app for freelance designers? | Assistant B | Yes | 3rd of 4 | Two competitors | Price wrong |
This is an example of the shape, with made-up data. The “facts right?” column matters as much as “named?”: being recommended with the wrong price is a problem you can fix.
4. Look for the pattern
After ten questions you will usually see one of a few patterns.
What the results mean
| What you see | Likely cause | What to do |
|---|---|---|
| Never named | AI cannot open or understand your pages, or does not know you exist in this market | Check crawler access and that the text is in the HTML, then say plainly what you are. See this post and this one |
| Named, but facts wrong | Your pages disagree with each other, or an old page or another site says something different | Make every page agree, especially prices. See making prices readable |
| Named, but after competitors | No concrete reason to choose you, or unanswered buyer questions | State specific advantages and answer the questions buyers ask. See the questions buyers ask |
| Named only sometimes | Normal variation; you are on the edge of the shortlist | Keep tracking; strengthen the pages and look at who is named instead |
Also write down who is named instead. Those are your real competitors in AI answers, and they are not always the ones you expect.
Repeat it, and measure the trend
One run is a snapshot. The same question can give a different answer tomorrow, and assistants change. Ask the same set once a week for a month before you draw conclusions, and look at the share of answers that name you rather than any single result. Keep the question list stable so the numbers compare.
What this method cannot tell you
- It shows what assistants said on the days you asked, not what every user sees. Answers can differ by location and settings.
- It cannot tell you why you were missed. For that you need to look at what the assistants can read on your site.
- Doing it by hand takes time. Ten questions on three assistants is thirty answers a week.
Where Unvibe fits
The manual method works, and costs only your time. Unvibe does the repetitive part.
AI mentions, rank tracking and fact check
- AI mentions: asks AI your buyer questions the way a normal user would and records whether you are recommended, your position and who is named instead.
- Weekly rank tracking: repeats the questions every week, so you see the trend.
- Fact check: asks an AI what it knows about your app, then checks each claim against your live site and gives a correction for every wrong or missing fact.
- Supporting evidence: when an AI skips you, the audit links the likely causes on your site. AI mentions, weekly tracking and fact check are on the paid plans.
See how these are measured, and their limits, in AI visibility and rank tracking and the methodology.
Questions
Can I use a free account to check?
Yes. The manual method needs only the free versions of the assistants. Be aware that free and paid versions, and different countries, can give different answers.
Why does the same question give different answers?
Assistants are not deterministic, and many search the web at the time you ask. That is why one result proves little and a weekly trend is more useful.
How many questions should I track?
Eight to ten is enough to see patterns without making the weekly check a chore. Keep them to the real questions buyers ask, and keep the list stable.