How AI-literate
is the Netherlands?

 

Prompt. Hallucination. Deepfake. Large Language Model. The language of AI has suddenly become part of everyday conversation. But how well do we actually speak this new language? Research among 1,000 people in the Netherlands shows that while AI use has become widespread, our understanding of the technology is struggling to keep pace.

People in the Netherlands are generally positive about their own AI knowledge. More than half of respondents say they understand and use the basics of AI. Among younger generations, that figure is even higher. 

The AI assessment included in our research paints a different picture. On average, respondents scored 5.1 out of 10: 4.9 for AI terminology and 5.4 for AI literacy, defined as the ability to understand and critically use AI. 

Take one example. Eight in ten respondents know what a deepfake is, yet only one in five can correctly explain what a Large Language Model actually is. And when it comes to neural networks, more than half draw a blank. The vocabulary is growing, but fluency is lagging behind. We increasingly recognise AI terminology without fully understanding what it means. 

AI has quickly established itself as part of daily life. People in the Netherlands primarily use AI to ask questions, find information, and write or improve text. 

Yet the research shows that many still have a mistaken view of how generative AI works. Just 14 per cent know that generative AI creates new content itself. Nearly half believe AI simply retrieves and summarises existing information, much like a search engine. 

That helps explain why AI feels familiar to many users. When it is mainly used to obtain answers, it can easily be perceived as a more intelligent version of Google. The risk, however, is that people overestimate how much they truly understand about the technology behind it. 

Perhaps the most striking finding from the research concerns trust. AI systems can make mistakes. They can even generate convincing but inaccurate information. In the AI world, this is known as a hallucination. Fewer than half of respondents know what that term means. 

The study also found that 17 per cent believe AI checks answers for factual accuracy before presenting them. A quarter think AI weighs up arguments in the same way a human would. Both assumptions are incorrect. 

This gap in understanding is reflected in behaviour. Only 16 per cent always verify AI-generated output against another source. Most do so only sometimes, rarely, or not at all. 

That presents a risk. If people do not understand how AI produces answers, they are less likely to recognise when those answers are incomplete, misleading or simply wrong. 

For many people, AI is still primarily associated with a chatbot on a screen. Behind the scenes, however, the technology is already playing a far greater role. AI is now used in a wide range of processes, including mortgage applications, recruitment and healthcare. Yet almost half of respondents are unaware of this. 

There is also considerable uncertainty about responsibility. When AI makes a mistake that affects an individual, many people are unclear about who should be held accountable. More than half either have an incorrect understanding or simply do not know. 

This suggests that AI is still widely viewed as a personal assistant, even as it increasingly becomes part of systems that support or influence decision-making. 

So, does the Netherlands already speak the language of AI fluently? Not yet. AI use is becoming more common, and people are growing familiar with the terminology. At the same time, knowledge of how AI works, its limitations and its risks continues to lag behind. 

That is not a cause for concern, but it is a signal worth paying attention to. As with any new language, understanding goes beyond recognising words. Only when we understand how AI works, when it can go wrong, and how to assess its outputs can we use the technology effectively and responsibly.