AI visibility: have it measured first.
More and more customers put their question to an assistant and get a handful of names back instead of ten links. For 75 euro we analyse your company name in the published AI Index series for your sector. You see at which questions you are and are not named, how that changes across seven stored moments and which names appear more often. The report uses published dataset version 2026.1 and does not start a new model run.
What you are actually buying
A private report on the existing public measurement series for your Dutch sector. We search your full company name, show every match by question and stored moment, put the most frequently named other companies beside it and explain the limits of the measurement.
Your order does not start a new model run. That keeps the source, version and result auditable and allows the same report to be reproduced safely.
Order the report → Online ordering is not open yet. The button appears automatically after payment, tax and private delivery have been released together.Available for these nine Dutch sectors
What a measurement produces is a series, not a score
Every question appears once in the table and holds seven answers side by side: the current one plus six stored earlier answers. So you see not only whether you get named today, but also whether you were still there four measurements ago. The example below really is in the question list for the Automotive sector; the names have been replaced because this only shows the mechanism.
the question as it is askedWhere can I order cheap car parts if I want to do small repairs myself?
Seven boxes, because there are seven measuring moments per question. A company named five times out of seven stands somewhere different from a company that was in it once. That series is the whole point: a single measurement says little, because a language model does not answer the same question the same way every time. How a measurement comes about is in the methodology of the AI Index, and every answer is in the open data.
Ten links became one answer with four names
A search result holds ten links and you choose yourself. An AI answer holds a short list of names, and anyone not on it is not considered. That changes what the fight is about. Sitting in eighth place at Google still brings clicks; in an answer of four names, eighth place is the same as not existing.
For a local business it is almost never about the question who is the best cleaning company in the country. It is about the questions a customer really asks: which company can do this near me, roughly what does this cost, what should I watch out for. What an assistant answers there can largely be traced back to what can be found about you and how unambiguous it is.
We have written that mechanism out in three places, and those stand apart from this service. Read what AI Overviews do to a trade business if you run one, how zero-click search changes your traffic if you want to know where the clicks go, and how customers use ChatGPT to look for a tradesperson for the customer's side.
Six of the ten places have no counterpart on the right, and eighth place is one of them
The paragraph above names two numbers: ten links in a search result and four names in an AI answer. In running text that stays a comparison; side by side it is a map on which you can point at where your place goes. This is the shape of the two results and not a measurement of your positions.
The change sits in the six striped rows. Anyone in fifth to tenth place at Google still has a position to work from; in an answer of four names that same business falls out of view without anything about it having changed. So the question on this page is not how high you sit but whether you get named, and that is why the outcome of a measurement is a series of yes or no rather than a position.
How we measure it
We do not guess and we do not ask a panel. We put the question to the model itself and keep the answer. Those same systems may use our own public content in return, on the two conditions set out in our AI policy. This is the whole method, and it has run since March 2026 as the AI Index.
There is a question list per sector. Those are not keywords but whole sentences, the way somebody would put them to an assistant, and that list is handwork: a cleaning company gets different questions from a car garage. At the moment nine sectors are running with 360 questions between them. Per question the model gets an instruction that does not change: name ten Dutch companies, names only, no explanation. That word Dutch is not a detail: the index that runs today is a measurement of the Dutch market, and what it says about the British or Irish market is nothing.
A question goes again as soon as the previous answer is older than 25 days. So there is no measuring day on which everything runs at once; every question has its own clock. When a new answer arrives, the old one shifts a column along and the seventh disappears. That keeps seven measuring moments beside each other per question without the file growing. Together that is 2,520 recorded answers at the moment.
We are questioning a model here, not the market. Today that is gpt-5-mini from OpenAI, through the API. That means we say nothing about Gemini or Perplexity, even though you might expect that of an index like this. We write that down rather than leaving it out, and the full account is in the methodology.
The whole dataset is open. Every question, every measuring moment and every answer can be downloaded as CSV and JSON under CC BY 4.0 through the open data, with the rankings per sector beside them, so you can check every count yourself. A measurement you are not allowed to check is a claim.
Three multiplications, and a window that explains why the number does not keep growing
The paragraphs above name four numbers in four places: nine sectors, 360 questions, seven measuring moments and 2,520 recorded answers. Separately they look like loose facts. As a chain you can see it is a sum, and which link you would have to change if you wanted to know what happens with a tenth sector.
The window shifts, the file does not grow. As soon as an answer is older than 25 days the question goes to the model again. The new answer comes in on the left, everything shifts along a place and the seventh drops off. So 2,520 is not a running total that climbs with time but a fixed width: nine sectors times forty questions times seven stored answers. A tenth sector changes the first link and with it the final number; a longer memory would change the fourth link, and we deliberately do not do that.
Ten places filling up, and then a question that waits
The paragraph above describes what happens per question: the model gets an instruction that does not change, gives ten names back, and that answer is stored. Below that happens too, again and again, because that is what separates an index from a snapshot. There are deliberately no names in the boxes: which companies the model names is a measurement, and that belongs in the open data and not in a drawing.
the instruction that does not changeName ten Dutch companies, names only, no explanation.
Ten places, because that is the instruction. The boxes stand for the ten names the model has to give per question, and the reading beside them counts how many are there. After that what the pill says happens: nothing, until this answer is 25 days old. All 360 questions work exactly like this, each with its own clock, so there is no measuring day on which everything runs at once. Anyone with motion switched off sees where this ends up: a full answer of ten names with the standby state beside it. What actually came out of those answers is in the open data.
What you get back after a measurement
The report analyses the already published question list and measurement series for your sector. It contains these parts.
- In how many answers you appear. A count across the whole question list, not a score from nought to a hundred. There is no standard to measure that against, so we do not invent one.
- In which questions exactly. This is the most usable part. If you get named for emergencies and not for maintenance, you know where the work is.
- How your mentions move across the series. Per stored moment you see how many valid answers name your company.
- Which other companies appear most often. The ten most frequently named other companies, including total and current mentions.
- Where the measurement ends. The report names the model, dataset version, invalid rows and five limitations. It makes no quality judgement and gives no guarantee.
If you want an impression first without an appointment, use the free AI brand checker. It puts your company name, sector and town to a language model and shows straight away whether it recognises you. It asks a single question and not a series, so treat it as a first impression and not as a measurement.
One answer against 280, on the same scale
The last paragraph above warns that the free checker gives an impression and not a measurement. On the same scale you can see how large that difference is: the checker asks one question at one moment, the measurement puts the whole question list of your sector at seven moments. Underneath it says per item from the list above which of the two can produce it.
The checker is the doorbell, not the measurement. It is there so you can tell within a minute whether this subject matters to you, and it costs nothing. Everything that asks for an order, a trend or a comparison needs more than one answer. Anyone who reads the checker as a result is reading a sample of one as a final standing.
What we do afterwards to improve it
A measurement is the beginning. The work after it looks a lot like good SEO, with a different aim: not climbing a list, but becoming recognisable and repeatable for a system that summarises.
Making sure you are one unambiguous entity. The same company name, the same address, the same services on your site, your Google Business Profile and the other places you appear. Anyone known under three spellings is three half companies to a model.
Answering the question that is asked. The question list shows which sentences really come in. A page that answers exactly that question, with a price indication, a working area and an example, is more usable than ten pages writing around it. How such a page is built, from the definition at the top to the FAQ at the bottom, is written out step by step in how to make content for ChatGPT, Perplexity and Gemini. And if you use AI to write it, the question of what Google does with that is answered in how Google reads AI content.
Adding evidence of your own. A figure you measured yourself, a method you describe yourself, an example from your own practice. That is the part that cannot be taken from another source, and that is exactly why it stands out.
What we do not promise
We promise no place in an answer, because nobody hands those out. We also promise no period within which an improvement comes through, because a model is not retrained at our request.
And then llms.txt. We have a free generator for it and that stays, because there are systems that read such files. But it is not a Google ranking factor: Google writes in its own AI optimization guide that Search ignores these files and that they do not affect visibility positively or negatively. Anyone selling it to you as a ranking service is selling air. What the file does do, who actually reads it and what belongs in it, is set out in llms.txt explained. The other side of the same question, keeping your work out of the systems you do not want it in, sits in the tool protecting SEO traffic from AI.
What it costs
All amounts are also on the rates page and exclude tax. A business in the United Kingdom or Ireland accounts for the VAT itself under the reverse charge.
- The AI brand checker: free. A single question to a language model about your company name, sector and town. No invoice, no trial subscription.
- A private AI Index report: 75 euro once. Analysis of the published dataset version for your sector, with the sections in DOC.03.
- Keeping track: 75 euro a month. With no annual contract. This subscription is not open for online ordering yet.
- With an ongoing SEO engagement: included. If you buy SEO at 250 euro a month, this is part of it and you pay nothing extra for it.
Two of the four steps carry an amount, and that amount is the same twice
The four ways to start with this are listed above under each other, which takes the order out of it. Side by side the stair runs from a single free question to an ongoing engagement, with the amount per step beside it. The height of a step stands for that step and not for the amount: the fourth step is the largest one and costs nothing extra, because it sits inside a fee that is already there.
Four steps, all four with the amount that is on the rates page. The height is the step and not the euro, so anyone comparing the amounts reads the line and not the bar. The fourth step is at nought because the amount beside it is already being paid, not because the work is free. All amounts exclude tax.
Who this is not for
A measurement shows where you stand. It solves nothing if there is no material under your business to tell anyone about. Three situations in which we would say: do something else first.
You have no figures of your own and no story of your own. Anyone who says exactly the same as the rest of their sector gets treated by a summarising system exactly that way: as one of many. Then start by describing your own work, not by measuring.
Your site is not technically reachable. If there is a block in your robots.txt or your site only loads after a script, there is little to find about you. First run the free SEO scan; it checks a page on 24 points and finds this kind of thing immediately.
You expect a guarantee. We cannot give one and neither can anyone else. What we can do: show where you stand now, what is missing, and over a number of measurements whether it has changed.
Four edges of this instrument, with what does and does not come out of it
Every measurement has an edge. This one has four, and they stand here side by side rather than in a footnote. Anyone reading the left column sees what you are buying; anyone reading the right column knows what not to use it for.
These four edges come from our own account of the method and not from a sales text. They are here because we ran into them ourselves while setting the index up; the full working out, including the answers that do not count, is in the methodology.
Questions buyers ask
Do you measure my brand or my sector?
Both, but in that order. The question list belongs to the sector, because those are the questions customers ask. What we send back is about your name: in which answers it appears and in which it does not.
Why a list per sector and not my own keywords?
Because people do not give an assistant keywords but sentences. Anyone measuring on keywords is measuring something other than what happens. See also how searching for local businesses has changed.
My sector is not in it yet. Is it still possible?
Not through the fixed 75 euro report. That works only for the nine Dutch sectors listed above. A new sector or country needs its own question list and measurement series first; that is custom work and we scope it separately instead of pretending the running index covers it.
How quickly do I see a difference?
We do not know and we will not attach a period to it. What we do know: you only see it across several measurements, because a single answer varies too much to draw anything from.
Is this the same as SEO?
The work afterwards resembles it and overlaps in part. The difference is in the aim: with SEO you want a place in a list, here you want to be named in a summary. If you buy SEO from us, this measurement is included.
Can I check the figures myself?
Yes. The whole dataset is open data under CC BY 4.0. You can recalculate every count and you may reuse it as long as you name the source.
Start at the free check
The quickest first step costs nothing: the AI brand checker asks for your company name, sector and town and shows whether a language model recognises you. If you then want the complete published series for your Dutch sector analysed, the private report is the next step.
Would you rather speak to somebody first? Book a call. You will be talking to the person who built the index, not to an account manager who has to go and ask.
