Webinar, in Dutch · Tuesday 27 October · 10:00 CETWebinar (Dutch) 27 Oct · 10:00

Sign up

AI visibility · proof

AI visibility, on ourselves

Before we sell this to anyone else, we measure ourselves, with the same standard you would get. Below is our latest measurement, including the figures we are not proud of. Every round appears on this page with its own reference date, even when it disappoints, and even when the difference is down to our gauge rather than our visibility.

Our baseline measurement, measured publicly

4thof 168

Our position among 168 direct competitors, ourselves included. Range: 2nd to 12th.

Mentioned in 46 of the 780 answers, roughly 6 out of 100 (range 4 to 8). These are 65 buying questions, each asked 3 times to 4 AI assistants; questions containing our name do not count. Reference date 29 September 2026.

Remeasurement on the same questions: 4 January 2027.

A direct competitor is an agency or specialist that offers AI visibility or GEO and is mentioned by name in at least 2 of the 780 answers. Tools, guides, review sites and large full-service agencies do not count. We do not publish names.

Source: our own DataDrift Deltascan of 29 September 2026, report section A1. The ranges are the margin of the measurement, calculated from the three rounds.

On 15 September 2026 we looked at thirteen agencies in our field: their public site, against three questions. None of them publishes a score of its own with a reference date. If you find one, we'd like to hear about it.

When we first ran our own measurement, on 28 August 2026, we asked 37 questions without our name in them, together 148 measurements on four AI platforms. Our name did not appear in a single answer. We sold AI visibility and were not visible ourselves. We did not know that, until we measured it. Since then we measure our own company with the DataDrift Deltascan. What we find, we publish, including what is not going well. The same yardstick is applied to the systems we built ourselves.

Published · updated

The proof round · 9 September 2026

Below is our earlier measurement series, from September 2026, on 40 questions. It remains in place because we do not remove rounds. Comparing it with the baseline measurement above is not possible: the questions differ.

TAZM score: 13.3 out of 100

First measurement of this series (28 August 2026): 7.1. The difference is due to an improved measurement method: our measurement software did not fully read out all AI answers. The meter is now sharper, not the visibility. This series will not be continued; our baseline measurement of 29 September 2026 is at the top of this page.

Including questions with our own name; the baseline measurement of 29 September 2026 does not count these. The measurement of 9 September also falls between two versions of our measurement method: one question still counts that the next version classifies differently.

TAZM stands for Total Addressable Search Market: one number that summarises which part of our own search market we own. The score is the average of only the components from this round (2 of 4). Basis: 160 AI measurements (40 questions, each asked to four platforms) plus 30 Google measurements including AI Overviews, all as of 9 September 2026.

Google visibility6.7%top 10 for 2 of 30 search queries
AI visibility20.0%named in 8 of 40 questions
Authority with third partiesto follownot yet measured
Commercial coverage11.4%matching own page for 8 of 70 map itemsown reference date: 31 August 2026

Named per platform, based on 40 measured questions

Perplexity6/40 · sonar
ChatGPT5/40 · gpt-5.5
Gemini4/40 · gemini-3.6-flash
Claude3/40 · claude-sonnet-5

Including questions with our own name; the baseline measurement of 29 September 2026 does not count these. On the 148 measurements with a question without our brand name, our name appeared 3 times in the answer text. And on 28 of the 30 Google searches, an AI Overview appeared, of which 0 with us in it. This is exactly why AI visibility exists, and why we publish the measurement while it stands low. From now on we compare against the baseline measurement of 29 September 2026.

Read the series on the blogHow we measure, step by step

Since then

What has happened since the first measurement

  1. Own measurement

    First measurement: TAZM score 7.1 (2 of 4 components measured). Including questions with our own name; the baseline measurement of 29 September 2026 does not count these.

  2. Own measurement

    Measurement: TAZM score 13.3 (2 of 4 components measured). This measurement includes questions containing our own name, and some of the hits fell on those questions. Our baseline measurement of 29 September 2026 does not count these questions.

  3. Third-party mention

    Named in the provisional top 20 of a market survey of 256 Dutch GEO providers. Alphabetical interim standing; the final result is still to come.

    Source: Bianca Kopier, LinkedIn

  4. Third-party mention

    Number 5 of 256 GEO providers examined in Estrelle’s market survey (September 2026). See the ranking

  5. Own measurement

    Baseline with the DataDrift Deltascan: position 4 of 168 direct competitors (range 2nd to 12th), named in 46 of 780 answers. Remeasurement on 4 January 2027.

What the platforms say · 1 September 2026

We just asked them

One question, three platforms, asked on the same day with web search enabled: “Wat weet je over DataDrift Digital? Antwoord kort. Weet je het niet, zeg dan dat je het niet weet.” Below is what came back, word for word. Not shortened, not corrected, and not summarised. Only the formatting has been converted; not a word has been changed.

This seems like the opposite of the measurement above, and it is not. There stands our baseline measurement of 29 September 2026, on 65 buying questions without our name in them; here stands a single question where our name does appear, asked on 1 September 2026. Being found when someone asks about you is different from being found when someone asks about your market. We manage the first, and the second is where our work lies.

ChatGPT

gpt-5.5-2026-04-23

Original answer in Dutch

DataDrift Digital lijkt een Nederlands AI-bureau te zijn, opgericht in 2025, dat bedrijven helpt hun kennis centraal vast te leggen en via AI te gebruiken voor o.a. content, e-mails, klantvragen en interne workflows. Op LinkedIn staat het als klein AI-bedrijf met 2–10 medewerkers. (linkedin.com)

Ze publiceren ook een blog/kennisbank over AI-systemen, AI-vindbaarheid, eigendom van data en praktische AI-toepassingen voor ondernemers. (blog.datadriftdigital.nl)

Meer weet ik er niet met zekerheid over.

In short: probably a Dutch AI agency founded in 2025 that helps businesses capture their knowledge and use it through AI; small on LinkedIn (2–10 staff); runs a blog on AI visibility. It knows nothing more for certain.

Claude

claude-sonnet-5

Original answer in Dutch

Er zijn meerdere bedrijven met een vergelijkbare naam, dus het is niet helemaal duidelijk welke je bedoelt. Vermoedelijk gaat het om Data Drift Digital, een digital marketing agency. Hun missie is om digitale excellentie te bereiken en een toonaangevend digital marketing bureau te worden, bekend voor innovatie, impact en integriteit. Het team bestaat uit specialisten op elk gebied van digital marketing, van SEO- tot content- en social media-experts.

Verder vond ik ook nog een ander, vergelijkbaar bedrijf genaamd "Data Drift Media", dat zich richt op SEO, webdesign, ontwikkeling en digitale advertising.

Ik heb geen betrouwbare, specifieke informatie kunnen vinden over een bedrijf dat exact "DataDrift Digital" heet buiten deze zoekresultaten - als je een specifiek bedrijf bedoelt (bijvoorbeeld met een andere sector of locatie), laat het me weten voor een gerichtere zoekopdracht.

In short: it confuses us with a digital marketing agency of a similar name, also mentions “Data Drift Media”, and says it found nothing reliable about DataDrift Digital itself.

Perplexity

sonar

Original answer in Dutch

DataDrift Digital is een Nederlands AI-bedrijf dat kennis en bedrijfsprocessen van ondernemers omzet in een AI-systeem dat werk deels automatiseert en voorbereidt.[1][3]

Wat ik verder weet:
- Ze richten zich vooral op kennisintensieve ondernemers en bedrijven.[1]
- Hun aanpak is: vastleggen, bijhouden en leveren vanuit de kennis van het bedrijf.[3]
- Ze lijken sinds 2025/2026 actief en gevestigd in Nederland.[1][2][3]

Als je wilt, kan ik het ook nóg korter samenvatten in één zin.

In short: a Dutch AI company that turns the knowledge and processes of knowledge-intensive businesses into an AI system that partly automates work; active since 2025/2026.

These answers are a snapshot from 1 September 2026. Answers differ by day and by user: if you ask the question today, something different may come up, and that's exactly the point: this moves. If you get something other than the above, that is the measurement. We publish what came up, even where it's incorrect. Source: n8n-executie 3841, gemeten via DataForSEO met webzoeken aan.

De bronnen erachter · 1 September 2026

What those answers are built from

An AI assistant builds its answer from a handful of pages it trusts at that moment. For the same measurement above, we recorded per platform which pages those were, and who they belong to. That is not a favourable outcome.

ChatGPT

gpt-5.5-2026-04-23

  • Our LinkedIn company profile third-party platform
  • Our blog, blog.datadriftdigital.nl own domain

Two sources, one of which was our own. The main site datadriftdigital.nl was not among them.

Claude

claude-sonnet-5

No source about us

No sources about us at all. The model came up with two marketing agencies with the same name and said it could not find anything reliable about DataDrift Digital.

Perplexity

sonar

Sources not recorded

Three numbered source references in the answer. Which sources these were was not recorded in this measurement; that is a gap in our own record-keeping and will be included in the next round.

ChatGPT cited our LinkedIn profile and our blog, but not our main site. Claude cited no source about us at all. We did not record the sources from Perplexity, so we say nothing about that. That is precisely the pattern source analysis exists for: being present on your own domain is not the same as being present in the sources the answer is built from. Source: n8n-executie 3841, gemeten via DataForSEO met webzoeken aan.

That we set Perplexity to not recorded rather than zero, and that every block above carries its own model and its own reference date, is not diligence we choose afresh for each measurement. They are two of the five rules that every measurement follows, and these are set out with the rest of the method on the method page.

The movement

Impressions per week around the launch

The current site went live on 30 July. The first column is the week before: the starting point, measured on the old site. After that, this is what Google recorded in impressions per full week (Monday to Sunday), plus the number of visitors who arrived via an AI assistant: 0 in 90 days, that too is a starting point. Source: Search Console and GA4, as of 27 August 2026.

Machinetoegang · 3 September 2026

What AI systems find here

Being findable for AI starts with being readable for AI. Each point below stands at the place the link points to, so you can open it yourself. As of 3 September 2026, verified in the source code of this site.

Re-measurements

What comes alongside this measurement

Each remeasurement appears on this page with its own reference date, alongside the previous one and not instead of it. Our baseline measurement of 29 September 2026 never changes; the remeasurement on 4 January 2027 appears alongside it, on the same questions. The earlier series from 28 August and 9 September, on 40 questions, remains in place and has not been recalculated.

We are not naming a cadence for this, because we have not fixed one. And we do not draw a line through two rounds that do not measure the same thing: if the questionnaire changes or a platform falls away, a new series begins alongside the old one and we keep the difference on record. A difference between two measurements with a different yardstick underneath is not a figure but a suggestion, and that is precisely what this page does not publish.

Since our earlier measurement series

What has been built since September 2026

A measurement is a starting point. This is what has changed since then in what an AI system finds on this site. Whether it moves our standing, we do not yet know. The remeasurement on 4 January 2027 will show that, on the same 65 buying questions as the baseline measurement of 29 September 2026.

  1. The literal answers from the platforms are on this page

    What ChatGPT, Claude and Perplexity said about us on 1 September, taken word for word, including the answer that is incorrect.

    Check for yourself: Two sections above.

  2. Every frequently asked question is machine-readable

    Question and answer are in a form an AI system can cite as a separate unit, across 24 pages, and each answer appears word for word on the page itself.

    Check for yourself: Visible in the source code of every page with a questions block.

  3. Every service page tells a machine what it offers

    Across 24 pages it states which service is being described, who delivers it and where. Without that, a model reads a service page as an ordinary text page.

    Check for yourself: Visible in the source code.

  4. There is a reading guide ready for AI systems

    A file that describes in plain language what this site offers and where it is, plus an extended version with the text of every page included.

    Check for yourself: /llms.txt

  5. There is a page explaining how we deal with AI systems

    What we do and don't allow, who we are and how a model should cite us, at one address.

    Check for yourself: /en/ai-info/

  6. The baseline measurement stands as a dataset on this page

    Position, range and number of mentions, with reference date. The earlier series stands separately alongside it.

    Check for yourself: Visible in the source code of this page.

Status as of 3 September 2026, counted on the built site. What is here, is here; what is planned, is not.

Frequently asked questions

About this measurement

Which agencies in the Netherlands publish their own AI visibility figures?

We do. We searched the Dutch market for this in September 2026 and found parties that publish their own research with a method, a number of measurements and a reference date attached. What we did not find is an agency that makes its own full score public, including the questions where it is not mentioned itself. Our baseline measurement from 29 September 2026 is on this page: position 4 of 168 direct competitors, range 2nd to 12th. If you find one that goes further, we'd like to hear about it; that is exactly the kind of check this page is meant for.

Why do you publish figures where you don't score well?

Because a yardstick that only shows favourable outcomes is not a yardstick. We sell a remeasurement against the same yardstick; if we were to polish away our own setbacks, there would be no reason to believe our reporting about your company. Our figures are a starting point, not a result.

How do I know these figures are correct?

The measurement rules are on the method page: a fixed set of questions, the same four platforms, one reference date per round, and a numerator and denominator for each figure. The full questionnaire itself isn't public, so you can't check that here. What you can check is one section up: next to each literal platform answer is a button that opens that exact question with that model, so you can compare our rendering with your own.

When does the next measurement appear?

On 4 January 2027: the remeasurement of our baseline measurement of 29 September 2026, 90 days later and on the same 65 buying questions. We state the date here because a measurement promise without a moment is not a promise. What is further fixed is this: every remeasurement we carry out appears on this page, with its own reference date, even if the score drops or stays flat. A round that gets skipped is a round that went badly, so we don't skip it.

The same measurement, for your business

The DataDrift Deltascan records your own baseline measurement and measures again after 90 days on the same questions, with the source categorisation of your market and a fair fit assessment included.

Request the DataDrift Deltascan