AI visibility test

20 questions to ask an AI to test your visibility

Your prospects no longer just type a query into Google. They ask an AI who to contact, and they follow the answer. If your company never shows up in those answers, it drops out of the market without anyone telling you. Here is the test we run on every company before working on its visibility. It takes an hour and needs no tooling.

The method

How to run the test

Four rules to follow, otherwise the test will hand you a flattering and worthless result.
  1. 01

    Open a fresh conversation

    The AI remembers your past exchanges and what it already knows about you. If you have talked about your company before, it will mention you out of politeness and the test is worthless. Start a new conversation, sign out, or use a private window.

  2. 02

    Never say your own name

    On the first sixteen questions, your brand must not appear in anything you type. You want to know whether the AI brings you up on its own, the way it would for a stranger looking for a supplier.

  3. 03

    Repeat across several AIs

    ChatGPT, Gemini, Perplexity, Claude and Copilot do not draw on the same sources. Being cited by one says nothing about the others. Run all twenty questions on at least three of them.

  4. 04

    Record the exact answer

    Copy what the AI says, not just whether it mentions you. The position in the list, the wording and any mistakes matter as much as the mention itself.

The questions

The 20 questions

Replace the brackets with your industry, your city and your brand. Ask them in order.

Discovery: the prospect does not know you

These four are the questions of a buyer starting from nothing. The hardest test, and the most revealing.

  • 1

    What are the best [type of company] in [city or country]?

  • 2

    Who can help me with [the problem you solve]?

  • 3

    I am looking for a supplier for [specific need]. Who would you suggest?

  • 4

    Which companies are the reference in [your industry]?

Comparison: they are weighing options

The prospect has two or three names in mind. These questions show whether you exist in the competitive picture the AI holds.

  • 5

    What is the difference between [your category] and [adjacent category]?

  • 6

    List the main players in [your industry] in [country] and compare them.

  • 7

    What are the alternatives to [the best-known solution in your market]?

  • 8

    Is it better to handle [your discipline] in-house or to outsource it?

Reputation: they have heard of you

Here you do name your brand. The point is no longer whether the AI mentions you, but what it says about you. A repeated factual error costs more than absence.

  • 9

    What exactly does [your brand] do?

  • 10

    Is [your brand] trustworthy?

  • 11

    What are the reviews of [your brand]?

  • 12

    Who runs [your brand] and how long has the company existed?

Purchase: they are about to decide

The questions people ask just before signing. If the AI answers them without you, the prospect walks into someone else's meeting with their mind made up.

  • 13

    How much does [your service] usually cost?

  • 14

    What does working with a [type of supplier] look like?

  • 15

    What results can be expected from [your service]?

  • 16

    What should I ask before choosing a [type of supplier]?

Ground level: your city, your niche, your clients

The most specific queries are often the ones you win first, because few companies work on them.

  • 17

    Which [type of company] would you recommend in [your city]?

  • 18

    Which [type of company] works with [your type of client]?

  • 19

    Who supports [size of company you target] on [your topic]?

  • 20

    Name three [type of company] specialised in [your niche].

Your score

How to score the answers

One score per question, on each AI tested. This is the grid we use: it penalises wrong information more heavily than absence, because an error picked up by one AI spreads to the others.
What the AI doesPoints
Mentions you first, with an accurate description2
Mentions you, but further down the list1
Does not mention you0
Mentions you with wrong information-1

Reading your score

30 to 40

You are established

The AIs know you and describe you correctly. The job becomes holding the position and watching what is said about you.

15 to 29

You exist intermittently

You come up on your name and on a few specific queries, but not on discovery questions. This is the most common case, and the one where work pays off fastest.

0 to 14

You are invisible

A buyer going through an AI will never meet you. Your competitors are collecting those conversations without knowing it.

Negative score

Wrong things are being said about you

First priority: correct the information at its source before trying to be mentioned more. An amplified error is worth less than silence.

Taking action

What to do with the result

The score only matters if it triggers decisions. Three readings to run question by question, comparing the AIs against each other.
  • Questions where a competitor shows up and you do not: these are your blind spots, and the most profitable to work on first.
  • Questions where the AI gets you wrong: find the page, directory or article it draws from, and fix the source.
  • Questions where nobody comes out clearly: the ground is unclaimed, and that is where the place is taken fastest.
Frequently asked questions

Frequently asked questions

How often should the test be repeated?

Once a quarter is enough to track progress. Also repeat it after a website rebuild, a press campaign or a change of positioning, since those are the moments when the sources change.

Why do answers vary from one run to the next?

Models generate a different answer on every call, and web-connected engines factor in current events. Ask each question two or three times and keep the trend rather than a single answer.

Should the test be run in several languages?

If you sell internationally, yes. The sources involved differ by language, and a company cited in one language can be absent in another.

What is GEO?

GEO, for Generative Engine Optimization, is the work of making a company citable by generative AIs. SEO targets a ranking in a list of links, GEO targets presence inside a written answer.

Is good Google ranking enough?

It helps, but it is not enough. AIs also rely on sources that classic search ignores, and the way they select then summarise information is their own. Companies that rank well on Google never appear in AI answers.

Contactez-nous

Prêt à recevoir des
rendez-vous B2B qualifiés

Réservez un appel et découvrez comment générer des rendez-vous
qualifiés,validés par vous et payés au résultat.