Students edition
PISA measures what AI is doing to reading, ministers get caught using chatbot speeches, and the thesis itself is up for debate.
PISA: a quarter of pupils use a chatbot weekly, reading skills decline
The research shows exactly which skill you should not let slip right now: critical reading.
There is little news about your field itself this week. That is why this edition is shorter.
The OECD presented the 2025 PISA study on 8 September. PISA measures what 15-year-olds can do every few years. The result for the Netherlands is gloomy: average reading skill has fallen further since 2022 and sits clearly below the EU average. At the same time, about a quarter of Dutch pupils use an AI chatbot weekly to look something up, write, or summarise. More than a third use AI weekly as a learning tool, similar to the average across fourteen EU countries. The study also shows where the drop is: in judging information and thinking critically about a text. The share of 'fast readers', pupils who skim texts quickly, has almost doubled between 2018 and 2025. That is exactly the skill you need to notice when a chatbot gets something wrong. This is about school pupils, not you. But they are the people who will soon sit next to you in lecture halls, and it describes a habit students share just as much. This week, read one academic article fully without a summary first, then write five sentences yourself about what it said. A chatbot always gives a smooth, finished answer. That reads nicely, so you check less. If you get used to ready-made summaries, you stop practising hard reading work: reading a tough paragraph twice, weighing a source, spotting contradictions. That skill fades if you don't use it, and you only notice when you should have spotted a mistake in an AI answer and missed it. There is also a gap in the study you should know about. According to PISA, pupils from wealthier families more often get the chance to develop AI literacy. Everyone uses the tools, but not everyone learns to judge them. That could actually widen the digital divide around AI. For you, this means: don't wait for your course to offer a class on it. The skill of distrusting an AI answer is now worth more than the skill of asking for one.
Background for subscribersThis edition shows that reading skill is under pressure, while AI use keeps growing. So practise deliberately with tough texts, without a summary first. It also turns out that detection tools like Pangram only give estimates, and that courses are increasingly asking for proof of your process: drafts, notes, and a log of your AI use. How you handle confidential information matters too: never type data from your placement or side job into a chatbot without permission or anonymising it. Finally: choose your AI tool deliberately, not automatically the newest or priciest one, since that isn't always right for ordinary study work.
Ministers caught using AI text in their speeches
Nieuwsuur and de Volkskrant found traces of AI text in speeches by, among others, Prime Minister Rob Jetten, Education Minister Rianne Letschert, Minister Dilan Yesilgöz, and State Secretary Derk Boswijk. Some speeches reportedly consisted almost entirely of chatbot text. For Letschert, it involved a speech in the main hall of TU Delft. Nieuwsuur used the detection tool Pangram, which flagged about fifteen percent of that speech as AI text. The ministry admits it sometimes uses artificial intelligence. That's striking: the minister in charge of your education is being judged with the same kind of detection tool that can scan your thesis. Keep this example handy for the conversation with your course about what a detection score is actually worth. Such a tool doesn't give proof, only an estimate. It looks for patterns typical of text generated by a model, such as how predictable the next word is. Hence a result like 'fifteen percent of this speech': a probability about a piece of text, not a verdict per sentence. Here's the tension for you. If a score of fifteen percent is enough to put a minister in the news, what does such a score mean for your paper? The tool can't tell whether someone used AI to organise ideas, to translate, or to write everything. That's why courses are shifting towards proof of process and disclosure. Make sure you can show how you used AI, then you won't need to argue over a percentage.
Background for subscribersThis is a taster.
Subscribers read the whole Students edition: every story, each with its background in plain language.
Subscribe →Not ready for that? Read the free letter first →
AI and tech, made simple.
This is the Students edition. Subscribers get the background to every story, in plain language.
Subscribe
inklaretaal · Amsterdam · © 2026