Does AI Help Children Learn, or Weaken Their Skills?

AI can help children learn faster, or it can quietly take over the practice that builds their skills. Which one happens depends on design and habits: research so far suggests that AI which hands children finished answers weakens learning, while AI that makes them try first can support it.

This article is for parents and educators. It sets out the case for AI as a learning aid, the evidence of harm, the open questions about social development, and simple ways to tip the balance.

What is the case for AI in children’s learning?

The promise is a patient tutor for every child. An AI can explain a step again in different words, give practice at the right level, and answer the hundredth “why?” without tiring. For families who cannot afford private tutoring, that could narrow a real gap. Analysis: the benefit is in more practice and faster feedback, not in fewer questions for the child to answer.

Can AI weaken children’s skills?

Yes, when it removes the effort. Three recent studies point in the same direction, though none is about young children specifically.

• Practice versus learning. In a field experiment with nearly 1,000 high school maths students, Bastani and colleagues found that students with an unrestricted chatbot did 48% better in practice but 17% worse on a later exam without it. A tutor version that gave hints avoided the drop.
• Offloading and critical thinking. A study of 666 people by Michael Gerlich found that frequent AI use was associated with lower critical thinking, through cognitive offloading. Younger participants were more dependent on AI. This is a correlation, not proof of cause.
• Engagement while writing. A small preprint, Your Brain on ChatGPT, recorded the brain activity of 54 people writing essays. Those using an AI assistant showed the weakest connectivity and the lowest sense of owning their work.

The pattern: when AI does the thinking, people perform better in the moment and learn less.

Can AI affect children’s social development?

This is the least understood area, and the one with the most worrying signs. AI companions, chatbots designed to act as friends or partners, are already common among teenagers: Common Sense Media found that nearly three in four teens have used one and half use them regularly. After testing popular companion apps, it concluded in 2025 that they pose unacceptable risks and should not be used by minors.

Analysis: children learn empathy and conflict resolution from relationships that push back. An AI designed to agree and keep them engaged offers neither. There is not yet long-term research on children, which is a reason for caution, not reassurance.

Are younger children more at risk?

Probably, though direct evidence is thin. The studies above involved teenagers and adults, and Gerlich found that younger participants were the most dependent on AI. Younger children are still building the basic skills of reading, counting, remembering and explaining, so there is more for an AI to take over and less experience to judge its answers. General chatbots already set a minimum age of 13, and tools for younger children need to be designed for their stage: shorter, simpler help that leaves the child to finish the task.

Analysis: the safest default for a young child is AI used together with an adult, for a purpose, and for a short time.

How can parents and teachers tip the balance?

1. Try first, then ask. Make it a family or class rule that children attempt a problem before asking AI.
2. Choose tools that give hints. Prefer AI that asks questions and gives the next step, not the final answer.
3. Talk about what AI is. Explain that it is a program that predicts words, can be wrong, and is not a friend.
4. Keep companions out. Follow Common Sense Media’s advice: no AI companion apps for under-18s.
5. Respect age limits. General chatbots such as ChatGPT require users to be at least 13, with parental permission under 18; younger children should use tools built for them, with an adult nearby.
6. Keep human time first. Use AI as one learning tool among many, never as a substitute for reading together, play and conversation.

Clarity Kids, our learning companion for children aged 6 to 12, has configurable depth and screens every interaction with a multi-stage safety filter. For a full checklist, see Child-Safe AI: A Framework for Learning and Privacy.

Frequently asked questions

Can AI make children worse at thinking for themselves?

It can if it does their thinking for them. Studies link heavy reliance on AI to weaker learning and lower critical thinking, especially when the AI gives finished answers. Tools that give hints and make children try first avoid much of that harm.

What is cognitive offloading?

It means letting a tool do mental work you would otherwise do yourself, such as remembering, calculating or organising ideas. Some offloading is useful; too much can mean the skill is never practised.

Can AI affect children’s social development?

There are concerns, mainly about AI companions designed for conversation and relationships. Common Sense Media advises that they should not be used by anyone under 18.

At what age should children start learning about AI?

Simple ideas, such as that AI is a program and can be wrong, can be taught as soon as children meet it, with an adult. Independent use of general chatbots is set at 13 and over by providers such as OpenAI.

References

1. Bastani et al. (2025), Generative AI without guardrails can harm learning: Evidence from high school mathematics, PNAS — An unrestricted chatbot improved practice but harmed exam results; a tutor with safeguards largely did not.
2. Knowledge at Wharton, Without Guardrails, Generative AI Can Harm Education — Practice scores 48% higher and exam scores 17% lower for the unrestricted group.
3. Gerlich (2025), AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking, Societies 15(1) — Frequent AI use correlated with lower critical thinking through cognitive offloading; younger participants were more dependent on AI.
4. Kosmyna et al. (2025), Your Brain on ChatGPT (preprint) — Weakest brain connectivity and lowest sense of ownership in the group writing essays with an AI assistant; 54 participants.
5. Common Sense Media (2025), Talk, Trust, and Trade-Offs: How and Why Teens Use AI Companions — Nearly three in four teens have used AI companions; half use them regularly.
6. Common Sense Media (2025), AI Companions Decoded — Social AI companions should not be used by minors.
7. OpenAI, Terms of Use — Minimum age 13; parental permission under 18.