ChatGPT has grown from a technical curiosity to an indispensable tool for millions of Dutch users in just two years. With 900 million weekly active users globally and over 50 million paying subscribers, it remains the fastest-growing application in internet history. But while the Netherlands enthusiastically adopts ChatGPT, the country is simultaneously building its own alternative: GPT-NL.
ChatGPT in the Netherlands: the numbers
Dutch adoption of ChatGPT has been remarkable. Where in 2024 the tool was primarily used by tech professionals and students, by 2026 it has penetrated virtually every sector. According to recent research, 42 percent of Dutch ChatGPT users employ the tool for work and study, but usage is shifting: increasingly, Dutch users turn to ChatGPT for holiday planning, recipes, and everyday decisions.
At the same time, the Omnichannel Index 2026 reveals that Dutch retailers are strong in marketing but lag behind in AI adoption. This represents an enormous opportunity for businesses that invest in AI tools now.
ChatGPT’s market share has declined from 87 percent in early 2025 to 68 percent in early 2026. On mobile, the drop is even steeper: from 69 to 45 percent. Competitors like Claude, Gemini, Grok, and DeepSeek are all growing faster. But in absolute numbers, ChatGPT remains unmatched, processing over 5 billion monthly website visits.
GPT-NL: the Netherlands builds its own language model
While the rest of Europe still debates AI sovereignty, the Netherlands is already building. GPT-NL is a national language model developed by TNO, the Netherlands Forensic Institute (NFI), and SURF. The project has been running its first pilots since late February 2026.
Five organisations are currently testing the model in practice:
-
Gem: a virtual assistant used by nearly 30 Dutch municipalities. In 2024, Gem handled close to 70,000 conversations with citizens. The pilot is investigating whether GPT-NL delivers better answers than currently deployed commercial models.
-
HIP (Clear, Intelligent, Productive): a communication assistant that helps civil servants draft government letters in plain language. Think letters about debt and benefits that are currently unnecessarily complex for citizens to understand.
-
NFI: the model is being fine-tuned on forensic data to improve classification in criminal investigations. When dealing with terabytes of evidence, processing speed and precision have direct consequences for proceedings.
-
TNO itself: testing GPT-NL for classified and privacy-sensitive research projects. The policy is “Copilot, unless” meaning commercial AI tools are the default, but GPT-NL steps in where data sensitivity demands it.
The number of pilots is set to grow to 10 by spring 2026, with a broader commercial rollout planned for the second half of the year.
Why GPT-NL matters
The importance of GPT-NL goes beyond national pride. There are three practical reasons why this project is relevant for Dutch businesses:
1. Data sovereignty
Every prompt you send to ChatGPT travels to OpenAI’s servers in the US. For many government institutions and businesses handling sensitive data, that is a problem. GPT-NL runs on-premise, on Dutch or European infrastructure. Your data never leaves the country.
2. Dutch language expertise
ChatGPT speaks decent Dutch, but it is trained on predominantly English data. GPT-NL is specifically trained on Dutch texts and already outperforms older models like GPT-3 on summarisation tasks. For tasks requiring deep understanding of the Dutch language and culture, such as legal texts, government communication, or dialects, a Dutch model can make a significant difference.
3. EU AI Act compliance
The EU AI Act imposes strict requirements on transparency and accountability of AI systems. A European model built from the ground up with these rules in mind has a head start over American models that need to be retrofitted.
ChatGPT vs GPT-NL: which should you choose?
For most Dutch businesses, it is not an either-or choice but a both-and strategy:
| ChatGPT | GPT-NL | |
|---|---|---|
| Best for | General tasks, creative work, code | Sensitive data, government, legal |
| Language | Excellent English, good Dutch | Optimised for Dutch |
| Data | US servers (OpenAI) | On-premise / EU infrastructure |
| Cost | From $20/month (Plus) | Not yet determined |
| Availability | Available now | Pilots Q1 2026, broad Q3-Q4 2026 |
| EU AI Act | Being adapted | Natively compliant |
What this means for e-commerce
For webshop owners and e-commerce entrepreneurs, this opens two interesting doors:
Content in Dutch: GPT-NL could eventually generate better product descriptions, FAQs, and customer service responses that sound genuinely natural in Dutch, without the sometimes “translated” tone ChatGPT can have.
Privacy-friendly personalisation: With a model running on your own infrastructure, you can safely use customer data for personalised recommendations without worrying about GDPR compliance of American cloud services.
At Virge.io, we are already building tools that leverage AI-generated content for e-commerce. We are closely following the development of GPT-NL because it has the potential to elevate the quality of Dutch-language product content to a new level.
The future of AI in the Netherlands
The Netherlands is positioning itself as one of the few countries in the world actively investing in its own language model. Not to beat the Americans, but to offer an alternative that aligns with European values around privacy, transparency, and sovereignty.
The coming months will be crucial. If the pilots succeed and the commercial rollout gains traction in H2 2026, GPT-NL could become a blueprint for other European countries. GLBNXT, a Dutch startup, earlier this year launched Europe’s first certified AI platform that operates entirely beyond the reach of US cloud law.
For Dutch businesses, the advice is clear: use ChatGPT now for what it does well, but keep GPT-NL on your radar. Chances are that within a year, you will need a hybrid AI strategy with commercial models for general work and a European model for everything sensitive.
The AI train is moving. The Netherlands is not just riding it, it is building its own locomotive.