AI for Kids: Tool or Substitute for Thinking?
Last year, a parent sent me a message that I have been thinking about ever since.
Her eleven-year-old son had been using an AI assistant to help with a history assignment. The work he submitted was good — well-structured, detailed, clearly written. His teacher was pleased. His mother was not, because she had watched the whole process from the kitchen table. He had typed a question. The AI had answered it. He had copied the answer. The entire interaction had taken four minutes.
"He learned nothing," she wrote. "But I have no idea what to do about it."
I have been teaching children aged 10 to 15 for more than twenty years, and I want to tell you that this mother's instinct was exactly right — and that the question she is grappling with is one of the most important questions in education right now. Not just for her son. For every child who is growing up with a device in their hand that can produce a plausible answer to almost anything in under five seconds.
The question is not whether children should use AI. They will use it. They already do. The question is what happens to a child's thinking when they use it — and whether the adults in their lives are paying close enough attention to notice the difference.
What Twenty Years in the Classroom Tells Me About How Children Actually Learn
Before I say anything about AI specifically, I want to share something I have observed consistently over two decades of working with primary and secondary school children: the moments that actually change how a child thinks are almost never the moments when they receive information. They are the moments when they struggle with something, cannot immediately resolve it, and have to keep going anyway.
A child who spends twenty minutes genuinely confused about a problem — turning it over, trying different angles, getting it wrong, adjusting — and eventually arrives at an understanding, has done something that cannot be replicated by reading the answer. Something has changed in how their brain relates to that type of problem. They have built a mental pathway through difficulty that will be there the next time a similar difficulty arrives.
A child who reads the answer has acquired information. That is not the same thing.
I am not describing a romantic view of struggle for its own sake. I am describing a neurological reality about how learning actually works — one that has been confirmed repeatedly by cognitive science research and that any experienced teacher will recognise immediately from their own observations. Difficulty, when it is manageable and supported, is not an obstacle to learning. It is the mechanism of learning.
This is why the arrival of AI in children's lives concerns me in a specific way. Not because AI is bad — I use it myself, every day, and find it genuinely useful. But because the particular thing AI does best is remove exactly the friction that makes learning happen.
The Four Minutes That Cost a Child Something He Does Not Know He Lost
Let me go back to that eleven-year-old with the history assignment.
In those four minutes, here is what did not happen. He did not have to search his memory for what he already knew about the topic. He did not have to organise his existing knowledge into a structure that revealed where the gaps were. He did not have to figure out which questions were the most important to answer. He did not have to find the words for what he wanted to say, which would have required him to understand it clearly enough to express it. He did not experience the productive frustration of knowing what he means but not quite how to say it — and then resolving that frustration through the act of writing.
None of that happened. AI handled it all. Efficiently and invisibly.
What the child produced looked like learning. It had the shape of learning — structure, detail, correct information. But the process that produces lasting understanding, that changes how a child relates to a topic and what they can do with it later, was entirely absent.
And this is the problem that most discussions about children and AI miss. They focus on outputs — is the work good? is it original? did the child cheat? — when the more important question is about process. What did the child's mind actually do? What capacity did this experience develop? What will be different about how this child thinks tomorrow, because of what they did today?
If the honest answer is "nothing," then the tool has not helped the child learn. It has helped them produce something that looks like the result of learning, without the learning happening.
What I Have Seen Happen to Children Who Use AI as a Crutch
I want to be specific about what this looks like over time, because the individual interactions seem harmless and the cumulative effect is not visible until it is quite significant.
Children who regularly use AI to generate first drafts, answer difficult questions, and resolve problems they have not yet seriously attempted themselves develop a specific pattern that I have now seen enough times to recognise clearly. When they encounter a genuinely difficult problem — one where AI cannot immediately be consulted, or where the AI's answer is not reliable enough to simply accept — they become unusually uncomfortable. Not just the normal discomfort of difficulty. A deeper discomfort, closer to panic, that comes from having lost practice with the experience of not knowing what to do and having to think your way through it.
Their tolerance for their own uncertainty is lower. Their persistence when something does not work immediately is shorter. Their confidence in their own thinking — in their ability to figure something out independently — is weaker than it should be for their age.
These children are not less intelligent. They are less practiced in the specific experience of using their own minds to work through difficulty. And that practice, accumulated over years, is what produces the kind of thinking ability that cannot be outsourced to any tool.
What Good AI Use Actually Looks Like for Children
I want to be clear: I am not arguing that children should not use AI. I am arguing about how they should use it, and what parents need to understand to guide that use effectively.
The difference between AI as a learning tool and AI as a thinking substitute comes down to one question: is the child's mind working harder or less hard as a result of this interaction?
When a child uses AI to check their own reasoning — "I think the answer is X because of Y and Z, does that make sense?" — their mind is working harder. They have already done the thinking. The AI is helping them evaluate it.
When a child uses AI to understand why something they got wrong was wrong — "I thought it was X but you're saying it's Y, can you explain why?" — their mind is working harder. The friction of being wrong is producing the engagement that makes learning possible.
When a child uses AI to explore a topic they are genuinely curious about further than their teacher or textbook goes — "I understand the basic concept, but I want to know what happens when..." — their mind is working harder. Their curiosity is driving the interaction, and they are building on existing understanding.
When a child writes a first draft themselves — even a rough, imperfect draft — and then uses AI to identify weaknesses or suggest improvements, both the writing and the evaluation of the suggestions are developing their thinking.
In all of these cases, the child is the thinker. AI is a tool in the service of their thinking. That relationship is worth preserving deliberately, because technology will always pull in the opposite direction — toward making things easier, faster, and requiring less effort. Easier and faster are not always the same as better for learning.
The Questions Parents Should Be Asking
Rather than trying to monitor every interaction their child has with AI — which is neither practical nor particularly productive — I encourage parents to develop the habit of asking a small number of questions that reveal whether AI is supporting or supplanting their child's thinking.
"Can you explain to me, in your own words, what you learned?"
This is the simplest and most revealing question. A child who has genuinely engaged with a topic can explain it to you, even imperfectly. A child who has processed an AI output without engaging with it cannot — they may be able to read it back, but the explanation in their own words will be hollow or absent.
"What did you think about this before you asked AI?"
This question matters because it establishes whether the child's own thinking happened at all. The habit of forming your own view before seeking external input is one of the most valuable intellectual habits a person can have — and it is one that AI, if used carelessly, systematically undermines.
"Where do you think the AI might be wrong here?"
AI makes mistakes. More importantly, AI presents everything with equal confidence regardless of how reliable the underlying information is. Teaching children to interrogate AI outputs — to ask where the reasoning might be weak, where the information might be outdated, where a different perspective might produce a different answer — is teaching them to think critically in exactly the conditions that matter most in an AI-saturated world.
"What decision did you make in this that AI could not make for you?"
This question is particularly useful for creative and entrepreneurial tasks. There is always a decision in any meaningful piece of work that required human judgment — a choice between approaches, a value judgment about what matters, a decision about who the work is for and what it is trying to do. Helping children identify and own those decisions reinforces that their thinking is irreplaceable, however much AI can assist with the surrounding work.
Why Entrepreneurship Education Is Particularly Valuable in an AI World
The work I do with children through KidStartupper is built around problems that AI can assist with but cannot solve for a child — because the solving requires human judgment at every stage.
When a child identifies a problem in their community that is worth addressing, AI can help them research it and find out what others have tried. But AI cannot tell them whether this particular problem matters enough to them to be worth their sustained attention and effort. That decision requires self-knowledge and values that are uniquely theirs.
When a child develops an idea for addressing that problem, AI can help evaluate it and suggest improvements. But AI cannot tell them whether the idea is genuinely original or simply recombining what already exists. That judgment requires creative confidence and a developed sense of what novelty looks like in a domain they care about.
When a child presents their idea to an audience, AI can help them prepare and anticipate questions. But AI cannot stand in front of the room for them, read the audience in real time, adjust to unexpected responses, or experience the specific feeling of having made another person genuinely understand and care about something they created. That experience, with all of its discomfort and satisfaction, is irreplaceable — and it is exactly the kind of experience that builds the confidence and communication capability that AI cannot produce.
The point of entrepreneurship education in an AI world is not to prepare children to compete with AI. It is to develop the specifically human capacities that AI makes more valuable: the ability to identify what matters, to communicate it compellingly, to make decisions with real consequences, and to take genuine responsibility for what you create.
What I Tell Parents Who Are Worried About This
When parents come to me worried about their child's relationship with AI, I tell them the same thing I have told parents worried about calculators, search engines, and smartphones: the technology itself is not the problem. The problem is always the same problem — a child who has learned to outsource thinking rather than develop it.
The solution has also always been the same. Children need enough experience of genuine intellectual effort — of working through difficulty, being uncertain, making mistakes, trying again, and eventually arriving at something they understood for themselves — that they develop confidence in their own thinking. That confidence is the thing that makes all the tools useful rather than substitutive. A child who trusts their own mind uses AI to extend it. A child who does not trust their own mind uses AI to replace it.
Building that confidence takes time, the right kinds of experiences, and adults who understand why the process matters as much as the product. It is one of the things I have spent twenty years learning how to do — and it is what KidStartupper is built around.
If you want your child to develop the thinking capacities that AI cannot replicate — through structured, project-based learning that puts their judgment, creativity and communication at the centre of everything — I invite you to try the programme. Seven days free. No commitment. No credit card required.
See what your child does when they have to think for themselves.
start your child's free 7-day trial here
You may also find these articles helpful:
the kids who will thrive in the AI era
why parents should stop solving every problem for their children
