Danil Sysenko · Netherlands
When an expat about to hire a lawyer or a tax adviser asks ChatGPT or Google’s AI answers who to contact, a handful of Dutch firms come up again and again. Most do not come up at all. I measure which one you are, and publish how I measured it.
On 22 September 2026 I put twenty questions an expat asks before choosing a lawyer to ChatGPT, Google’s AI Overview and Google’s AI Mode. Sixty runs, fifty-two of which produced an answer. Twenty-six Dutch immigration firms were named across them.
The whole thing is public: every question, every firm named, and a section on what the numbers do not prove. Read the measurement. The same was done for employment lawyers on 26 September, and for expat tax advisers in ChatGPT on 25 September.
That page is also the honest answer to “why you”. I am not asking you to trust a methodology I describe in a sales call. It is written down, with the file each number came from.
Two days after that measurement I ran the same twenty questions again, changing nothing and telling nobody to do anything.
On 16 of the 18 questions, a different set of firms came back. On Google’s AI Mode — the one surface I can compare honestly — eighteen firms were named on the first day and twenty on the second. Two dropped out, four appeared: 27% of the names rest on one of the two days and not the other.
This matters for two reasons, and the second one is the point.
First: a single snapshot describes a day, not a position. Mine included.
Second: anyone who tells you they moved your firm up owes you this number first. Without knowing how much the list moves on its own, an improvement and a coincidence look exactly alike. I have not seen another provider publish it.
Every statement carries the file, the date and the question it came from. If a fact about your firm is missing, the audit says it is missing rather than leaving a silence you might read as zero.
That I can put you at the top. I do not know that, and the drift figure is exactly why. What I can do is measure before, change one thing, and measure after — and report it either way.
That being named brings clients. I measure visibility in AI answers. Whether that visibility turns into instructions is a separate question, and I have not seen it demonstrated for Dutch legal work. If someone tells you they have, ask them for the measurement.
That the differences I find are causes. They are correlations. Some rest on a hundred firms and more; one of them rests on six, and the audit prints the sample size next to the number rather than under it. The report says so on every one of them.
Each price is the total you pay: nothing is added on top.
Your firm on the twenty questions a client asks, in ChatGPT and in Google’s two AI answers: where you come up, who comes up instead, and up to three changes I would make first. One page, within 48 hours of your reply. You pay after you have read it.
You make one change — a map-listing category, a page for a question your clients ask — and I measure whether it moved: the same questions before and about four weeks after, in all three AI answers, against a threshold I write down before the second measurement. You get both sets of answers, whether the AI picked up the change itself, and a verdict that says plainly when nothing moved beyond the normal day-to-day movement. You pay after you have read it.
If you would rather not write the page yourself, I write it: the page that answers the client question you are missing on, in your voice, from facts on your own site, with one round of changes after your comments. You check the facts and publish it. The same before and after measurement is included, and so is the check of whether the AI quotes the page. I invoice when the page is live. I cannot promise the answers will move; if they do not, the report says so, and the page is yours either way. All prices on one page.
Two measurements, not one: your firm across the twenty questions today, and again three weeks later, compared against how much the list moves on its own. Plus your site read page by page, and the written audit with the source file named next to every number in it.
The list rewrites itself: asked the same questions two days apart, AI Mode returned a different set of firms on 16 of the 18, with nobody touching anything. A position measured in September says little about November. So every month you get the same questions asked again, and a short report: where your firm was named, where it dropped out, who took the place, and whether the change is bigger than the list’s own movement or sits inside it.
Month to month, stop whenever you like. It can follow the audit or start on its own.
Start with the quick check. It tells you where you stand today and what I would change first. Then make one of those changes and test it, or have me write the page. Monitoring is for when you want to keep watching after that; the audit is for when you want your whole site read as well.
I have not done this for a paying client yet. The measurement on this site is the work; judge it and decide.
AI answers are not the only place a client’s attention goes, and they are not separate from the rest of search. In this measurement, only 34% of the domains Google’s AI Overview cited also appeared in the organic results on the same page. On three questions, not one cited domain was in the organic list.
Which is to say: an SEO report that calls your rankings healthy is describing a different system from the one answering your client’s question — and the money you put into Google Ads sits in a third system again, with its own blind spot. Blended ROAS hides what non-brand is really doing, per market.
I audit that too, and the two fit together: the same client, the same question, three different places where the answer is decided. The Google Ads audit is a separate piece of work with its own price and its own report.
One person, not an agency. I built the measurement, the crawler and the checks myself; the numbers on this site are printed by a tool that names its own source file for each of them, because I have twice caught myself repeating a figure that the data no longer supported.
That is the whole pitch: a method you can check, and someone who tells you when it did not work.
The simplest way to start: reply to my email, or write to danil@sysenko.com, and I will send your firm’s list the same day: every question, and who was named on it. It is free, and nothing follows unless you ask.