Pupils and chatbots: age limits and what you can do in class
Many chatbots are not meant for children under 13, and Claude only for adults. What that means for primary school, and how you can still teach pupils to use AI critically.
- ChatGPT requires a minimum age of 13, with parental permission under 18. Claude is for 18 and over. For Gemini and Copilot it depends on country and account type.
- Under the Dutch GDPR implementation, parental consent is needed for children under 16 when processing is based on consent.
- In class it works well when the teacher operates the chatbot on the interactive whiteboard and the class checks the answers together.
The age limits
Most well-known chatbots are not made for primary school pupils. ChatGPT's terms require a minimum age of 13, and parental permission under 18. Claude is only meant for people aged 18 and over. For Google Gemini and Microsoft Copilot it differs per country and per type of account, such as a school account or a private account. Always check the provider's current terms; they change.
The GDPR also applies: in the Netherlands, parental consent is needed for children under 16 when data is processed on the basis of consent. So do not let pupils create a chatbot account themselves.
What does work well in class
- You operate it, the class thinks along. Ask a question the class comes up with on the interactive whiteboard and discuss the answer together. Is it right? How do you know?
- Find the mistake. Have the chatbot write a text about a topic the class has just learned, and let pupils find the errors.
- Human or machine? Show two short texts, one by a child and one by AI, and discuss the differences.
That way pupils learn the most important thing: a chatbot is not a search engine or encyclopedia, but a text predictor that is sometimes convincingly wrong.
Pupils who already use AI at home
Many children already know chatbots, through a parent, sibling or an app. Do not assume they don't use them; make it something you can talk about. What do you ask it? What do you do if the answer is strange or frightening? And tell them who they can turn to. Fake images and deepfakes belong in this conversation too; in the Netherlands, children and parents can go to Helpwanted.nl about image abuse.
Projects and pupils' own work
An AI detector is not reliable enough to judge a child on. It is better to see the process: let pupils work on their project in class, ask them to explain what they learned, and agree when AI is allowed, for example to find ideas. Also read Lesson preparation with AI.