Dutch marketing teams should monitor six numbers to improve discovery in AI assistants: Visibility Score, Share of Voice, citations (and which sources they come from), sentiment, average position, and AI-referred traffic. Measure them on Dutch-language prompts, per AI model, on a fixed daily schedule, because AI answers are volatile and differ by language and model.
Most guides on this topic come from US vendors and assume an English prompt set and Google as the default. In the Netherlands, buyers ask in Dutch, compare against local players such as Bol.com, Coolblue or Thuiswinkel-certified shops, and increasingly meet your brand inside ChatGPT, Gemini, Perplexity or Google AI Mode before they ever reach your site. This guide turns the generic KPI lists into a measurement setup for a Dutch team.
Why the usual KPIs are not enough
Rankings, impressions and click-through rate describe a results page. An AI answer has no page two: your brand is either in the answer, in the cited sources, or absent. That changes what you need to count.
Two findings explain why a single snapshot misleads. AirOps' 2026 State of AI Search reports that only 30% of brands stay visible from one AI answer to the next, and 20% across five consecutive runs (AirOps). The same source says roughly half of cited pages change every month. Note that AirOps sells AI search tracking and does not publish its sample sizes, so treat these as directional. The practical lesson stands either way: you need trends over many runs, not one screenshot.
The six metrics to monitor
Metric | What it tells you | Typical Dutch use |
|---|---|---|
How often and how prominently you appear across tracked prompts | One headline number for the management dashboard | |
Share of Voice | Your mentions as a share of all brand mentions in your category | Benchmark against Dutch and international competitors |
Citations and cited sources | Which of your pages, and which third-party pages, models rely on | Decide where to publish and where to earn mentions |
Sentiment | Whether the model describes you positively, neutrally or negatively | Spot "kan geschikt zijn voor..." style lukewarm framing |
Where you sit in lists and recommendations | Track the move from fifth to second place | |
AI-referred traffic | Visits and conversions that come from AI assistants | Connect visibility to revenue |
The first five can only be measured by running prompts and reading the answers. The sixth comes from your own analytics, which is why it needs separate treatment (see below).
Metric 1 and 2: Visibility Score and Share of Voice
Visibility Score answers "are we there?" and Share of Voice answers "compared with whom?". Together they stop you from celebrating a rise in mentions while a competitor grows faster. If you track 20 prompts and a competitor appears in 12 of the answers while you appear in 5, your absolute number can go up and your position in the category can still get worse. Read our explanation of how Trackbase calculates the Visibility Score for the formula, and how to measure AI Share of Voice for the competitor side.
Metric 3: Citations, with the source behind them
A mention tells you the model knows you. A citation tells you which page convinced it. Track both: AirOps reports that brands earning both citations and mentions are 40% more likely to resurface across multiple answers than citation-only brands. For a Dutch team, the source view is where the work starts. It shows whether models lean on Dutch review sites, forums, news outlets or your own pages, and therefore whether the fix is on-site content, structured data or earning coverage elsewhere. See mentions vs. citations for the difference in detail.
Metric 4: Sentiment, measured in Dutch
Sentiment is the metric most often skipped and the one that most often decides who wins a recommendation. Dutch answers hedge in their own way, so test whether your tooling classifies Dutch phrasing correctly. Our guide to tracking sentiment in Dutch AI answers covers the prompt set and the Net Sentiment Score.
Metric 5: Average position
Position matters because lists are short. Being named is necessary; being named first or second is what drives clicks and shortlists. Track position per model, since Gemini and ChatGPT often order the same brands differently.
Metric 6: AI-referred traffic, with realistic expectations
Visibility is an input; traffic and conversions are the outcome. Adobe lists AI referral and "dark traffic" attribution and assisted conversions among its core AI search KPIs, and AirOps cites reports that LLM-referred leads convert at two to six times the rate of other sources (again vendor-sourced, so validate on your own data). Be aware that much AI influence shows up as direct traffic, so a flat GA4 line does not prove a flat AI line. We explain this in why you cannot measure AI search in Google Analytics.
How to set it up in a Dutch context
Write Dutch prompts. Use 20 to 30 prompts phrased the way a Dutch buyer types them, covering category, comparison and evaluation questions. Do not translate your English set.
Track the models your buyers use. At minimum ChatGPT, Gemini and Perplexity; add Google AI Overview and Google AI Mode if search is a major channel.
Run daily and keep wording stable. Rephrasing a prompt resets the trend.
Benchmark against 3 to 10 competitors, including the local players that Dutch models cite most.
Review weekly, decide monthly. Use weekly numbers to spot swings and monthly averages to judge progress.
A simple reporting rhythm
Give leadership one slide: Visibility Score, Share of Voice and sentiment against last month. Give the content team the source view and the prioritized fixes. Give analytics the referral and assisted-conversion numbers. Three audiences, three views, one underlying data set.
Where Trackbase fits
Trackbase is a Dutch, EU-hosted platform that runs your prompts daily across up to 11 AI models and reports Visibility Score, Share of Voice, mentions, citations, sentiment, average position and competitor rankings in one place, with prioritized actions on top. You can see where you stand today with the free AI Visibility Checker, compare plans on the pricing page (14-day free trial), or explore the features. Teams that start from a Dutch prompt set usually learn the most in the first two weeks, when the source view reveals which pages models actually trust.
FAQ
How many prompts should a Dutch marketing team track?
Start with 20 to 30 Dutch prompts that mirror real buyer questions across category, comparison and evaluation intents. That is enough to see trends without drowning in noise. Add more only when you open a new product line or market.
Which AI models matter most in the Netherlands?
ChatGPT, Gemini and Perplexity are the sensible minimum, with Google AI Overview and AI Mode added if organic search drives much of your pipeline. Check each model separately, because visibility differs by model rather than moving together.
Can I measure AI visibility with Google Analytics alone?
No. GA4 only sees visits that actually arrive, and many AI-influenced sessions appear as direct traffic. Use prompt-based tracking for visibility, citations and sentiment, and use GA4 for referral and conversion outcomes.




