Everything about AI visibility
AI visibility · guarantee
Do I get a guarantee that I will appear in AI answers?
No. Nobody can guarantee that your company will appear in ChatGPT answers, because those answers differ by day, by phrasing and by user, and no supplier has access to the model that generates them. What can be recorded is the measurement: we fix the question list in advance, measure again with the same method on the same questions, and publish the outcome even if it disappoints. Our baseline measurement of 29 September 2026: position 4 of 168 direct competitors, mentioned in 46 of 780 answers. That is published openly on this site.
Published · updated
Why it cannot be done
An AI answer is a sample, not a standing
The same question yields a different answer today than tomorrow, a different phrasing yields different sources, and two users do not see the same thing. Anyone promising a position in that is promising something they cannot even observe, let alone steer.
On top of that, nobody in the chain has access to the model. No setting, no submission form, no back door; there is only the public material a model draws on. Anyone who guarantees you a place is guaranteeing the outcome of a process they stand outside of.
That is why we never present a measurement as a live standing. It is not a caveat at the foot of the page, it is the whole argument.
What is fixed
What we do guarantee
Three things, and all three can be checked without taking our word for anything.
- 1The list of questions is set in advance. You know which questions will be measured before any measuring happens, so picking a more flattering question afterwards is not possible.
- 2The method stays the same. The same questions on the same platforms every month, with the model name and the reference date alongside. A series is only a series if nothing shifts in between.
- 3The outcome is published, even when it disappoints. We already do this on our own measurement, and it can be checked before you buy anything.
Three layers, three kinds of promise
Where the limit sits per layer
The work falls into three layers, and we can promise something different in each. That distinction is why we can be firm on one layer and not on another.
| Layer | What the work is | What we promise |
|---|---|---|
| 1Technical | Machine access and the measurement system: structured data, a sitemap that holds up, and pages that answer the question your customer actually asks. | The delivery. It is either there or it is not, and you can check it yourself. |
| 2Publishing answers | Publishing every month on the questions that come out of the measurement, then re-measuring whether that answer is found. | The rhythm and the quality rules. Not the positions, and not the number of mentions. |
| 3Being mentioned elsewhere | Trade media, listings, reviews and communities: measuring which sources feed your market, and guiding you in connecting to them. | Effort only. Nobody promises an outcome here, and anyone who does is buying mentions. |
The third layer is also our own weakest component, and that is simply stated in our measurement. It is where the real work sits, and it is exactly why we attach no result to it.
Our own figure
What our own measurement shows, including the part that hurts
We apply this to ourselves first. Our baseline measurement of 29 September 2026 is published openly, with figures that are not dressed up. Questions featuring our own name do not count towards this.
Someone else does it differently
There is a party that does promise money back
Estrelle returns 50 per cent of the investment if your site does not appear as a recommended source or expert answer in 5 named AI models within 90 days of completion, on questions agreed in advance. Read on their own page on 5 September 2026.
The guarantee on their own page
That is the sharpest commercial move in this field, and it is only possible because they measure. We measure too and still do not do it, for one reason: the test hangs on an outcome neither party controls. If it is achieved, nobody knows whether the work caused it; if it is not achieved, half comes back and you still do not know what does work.
So we attach the promise to the part we do control: the delivery, the rhythm and the measurement. That sells less easily, and it is why this page exists.
Frequently asked questions
What people ask about this
Can you guarantee that I will appear in ChatGPT?
No, and nobody can. AI answers differ by day, by phrasing and by user, and no supplier has access to the model that produces them. What we do fix is the measurement: the same list of questions, the same method, every month, and the outcome published even when it disappoints.
What happens if the measurement does not move for months?
Then we say so, and we advise stopping ourselves. The monthly amount pays for work and measurement, not for an outcome. If the series shows no movement, you should hear that from us rather than having to discover it yourself.
Do you give money back if there is no result?
No. There is a party in the Netherlands that does, and it is on this page with its source and reference date. We attach our promise to the delivery, the rhythm and the measurement, because those are the parts we actually control.
Why do you publish a figure you score badly on?
Because a measurement that only appears when it is flattering is not a measurement. We counted 10 Dutch providers and none of them published a figure about their own work. Anyone who does not show their own figure is asking you to take theirs on trust.
So what can actually be checked?
Three things: the list of questions is set in advance, the method stays the same from month to month, and every measurement carries its reference date and the model name. Our own series is public, so you can check the method before you buy anything.
Know where you stand first, without a promise up front
The DataDrift Deltascan records your baseline measurement, measures again after 90 days on the same questions, and is a fit check at the same time. If there turns out to be too little to gain, we say so and it stops there.