How to choose an AI visibility checking tool
An AI visibility checking tool sends real questions from your market to ChatGPT, Perplexity, Google AI Mode and Copilot, and counts how many answers mention your brand, in what position and which sources the engine cites. To choose one, these are the ten questions worth asking before you pay. They apply to us as much as to anyone else.
1. Does it publish its methodology?
If you cannot read how it measures, the number it hands you is not verifiable — and an unverifiable number cannot support a decision.
This is the first question because it filters almost everything else. A tool that explains its method can be audited; one that does not has to be taken on faith.
2. How many times does it measure each question?
If the answer is "once", you are being handed noise with two decimal places.
These engines return different answers to the same question on the same day. With a single pass there is no way to know whether a rise is real or the coin landing the other way up.
3. Does it give a confidence interval?
A percentage without a margin looks firmer than it is, and over 90 questions that margin is large.
Without it you will end up celebrating two-point gains that fit entirely inside the measurement error — and, worse, making content decisions based on them.
4. Does it exclude questions that already name your brand?
If it does not, your percentage is inflated by construction.
A question like "what is company X like?" gets answered by mentioning X almost every time. Counting those raises the figure without visibility having changed at all. It is the easiest and quietest way to flatter a report.
5. Does it measure in the real interface or only through the API?
They are not the same, and the gap can be dozens of points.
The same model answers differently depending on whether it searches the web or replies from memory. A tool that measures only through the API is telling you what the model remembers, not what your customer sees.
6. How many questions does it measure, and for how much?
Coverage is what makes the result useful or useless: twenty questions do not describe a sector.
Look at the price per question measured rather than per month, and check that coverage is not quietly trimmed when the provider's own costs bite.
7. Can you fix the competitor list?
And if you do, does what is already measured get recalculated, or only what comes next?
Nobody knows your competition better than you, and somebody is always missing from the initial list. If fixing it means waiting for the next run, a five-minute oversight becomes a lost month.
8. What happens to your data if you leave?
Ask whether you can export the full history, and in what format.
The value here is the time series. A tool that measures well but keeps your two years of history is locking you in through the wrong door.
9. Which engines does it cover, and on which plan?
Measuring one engine tells you almost nothing about the others: two AI search engines answering the same question share between 4% and 13% of their sources.
That figure comes from 9,124 pairs of answers measured by vilnux Research. Also ask which engines are included in the plan you are buying, not the top one: the list on the homepage is usually the top plan’s. Beyond that, which ones matters more than how many: the ones your customer uses.
10. In which language are the questions written?
A question translated from English is not the one your customer asks, and the engine answers what it is asked.
A translated interface says nothing about this. Ask to see the questions it would run for your sector before you buy: if they read like a translation, the measurement describes a market that is not yours.
About the effectiveness figures going around Almost every multiplier repeated in this industry — that one thing triples citations, that another multiplies them by some number — has no study behind it. The only published, peer-reviewed work measuring the effect of optimising for generative engines is Aggarwal et al., KDD 2024, and what it finds is that adding statistics, quotes and references improves visibility by up to 30-40% in relative terms on its own benchmark. For any figure rounder than that, ask for the source.
Frequently asked questions
What does an AI visibility checking tool do?
It sends real questions from your market to AI search engines such as ChatGPT, Perplexity, Google AI Mode and Copilot, and counts how many answers mention your brand, in what position, alongside which competitors and which sources the engine cites to answer.
What should you ask an AI visibility tool first?
Whether it publishes its methodology. If you cannot read how it measures, the number it hands you is not verifiable, and an unverifiable number cannot support a decision.
Why does it matter how many times each question is measured?
Because these engines return different answers to the same question on the same day. With a single pass there is no way to know whether a rise is real or the coin landing the other way up.
How does an AI visibility report get inflated?
By counting questions that already name the brand. They get answered by mentioning it almost every time, so they raise the percentage without visibility having changed. It is the easiest and quietest way to flatter the figure.
Is measuring through the API the same as the real interface?
No, and the gap can be dozens of points. The same model answers differently depending on whether it searches the web or replies from memory. API-only measurement reports what the model remembers, not what your customer sees.
How many engines should an AI visibility tool measure?
More than one: two AI search engines answering the same question share between 4% and 13% of their sources, so measuring one tells you almost nothing about the others. Beyond that, which ones matters more than how many, and it is worth checking which ones are included in the plan you are buying.
Does the language the questions are written in matter?
Yes. A question translated from English is not the one a Spanish-speaking customer asks, and the engine answers what it is asked. Before you buy, ask to see the questions that would be run for your sector.
Start by checking us
The free audit answers these questions with your own domain in front of you. No signup, no card.
Run my free audit →Home · Plans & pricing · Methodology · Research · Perplexity · Metrics · Use cases · Changelog · Agencies · Alternatives · Diagnosis · Check if AI cites you · Noise and margin of error · Four-step method · Optimizing for agents · X · Product Hunt