For years, access and the digital divide have been burning topics of debate, often leading to the assumption that access to information can help to democratize opportunity. However, what seems to be missing from the discussion has been that AI is sold to us as a democratiser of intelligence, but it may create a far more consequential divide.
Today, a single command can write the essay, solve the equation, summarise the book, generate presentations, and brainstorm the ideas, many times free of cost.
However, there is a larger question we are overlooking:
What happens when the technology provides answers before the child has learned how to think? What happens when children, whose intelligence is still being formed, start consuming the responses, even before they learn to question?
Before I am accused of being anti-technology and anti-progress, I want to clarify: this is not a piece against using AI as a tool to save time or to enhance quality. Rather, there is a broader argument here, because we are too enchanted by the cost and time effective aspect of AI: what happens when children use it not to extend their thinking, but to avoid doing the thinking themselves?
This distinction matters a lot, because children don’t just need information; they need to develop their cognitive muscles which help them make sense of that information.
If we millennials reflect on the non AI days, we will be able to recall that we as children learnt not merely by writing a good essay or cracking calculations. We learnt by staring at a blank page, struggling to think of a topic, forming an argument endless times, choosing the right words, discovering contradictions and sometimes dismissing the entire idea and rethinking alternates. Analytical ability isn’t developed by receiving ready-made answers, but rather by problem solving, observation, listening and wrestling with many bad ideas.
The critical point of concern is
what happens to the mental muscles that were supposed to develop while the child struggles for answers? The real danger here is not that AI will give wrong answers. It is that it may give children answers before they have learned how to think.
The more we replace cognitive work with AI, the more dependent we will become.
An easy example can be how incapacitated we feel while navigating without Google Maps, or when the internet goes off. Now consider a generation who need support for logical thinking.
The evidence isn't simply “AI is bad.” Recent OECD/PISA findings suggest that how AI is being used matters: students using AI to replace cognitive effort tended to perform worse, while students trained to critically evaluate AI-generated material performed better.
Microsoft Research's study of 319 knowledge workers showed higher confidence in AI was associated with less critical thinking.
OECD (2026), PISA 2025 Results (Volume I): Future-Ready Students, PISA, OECD Publishing, Paris, https://doi.org/10.1787/73451bc5-en.
The productivity paradox
For adults, the AI argument is more compelling. I feel ChatGPT and Claude can be efficient time saver tools, which enable me to spend my time on higher-value work, whilst outsourcing the summarizing of long documents or low stake work. However, over time I have realised that overusing AI was diminishing my cognitive ability. I was drafting all email replies over ChatGPT, and after many months, when I wanted to write an email myself, I found myself at a loss for words.
The logic becomes more complicated when applied to children. The hours spent on solving a problem or writing a thesis are not just about creating an output; it’s a process of learning to construct and deconstruct an argument, to critically analyse various facets of the problem, to explore all possibilities with a human lens, and arrive at a conclusion. Learning is a journey, and having no answers and getting frustrated at not arriving at a solution are its essential elements. AI changes this equation because it can remove the friction almost instantly. And friction, inconveniently, is often where learning happens.
A 2025 randomized controlled trial compared 120 undergraduates studying with ChatGPT versus traditional study methods. After 45 days, the AI-assisted group scored 57.5% on the retention test versus 68.5% for the traditional-learning group. The researchers attributed this to reduced cognitive effort and cognitive offloading.
If cognitive effort is important for learning and critical thinking in older students, the question becomes particularly consequential when AI is introduced during childhood, when those cognitive capacities are still developing. In other words, AI isn't inherently damaging to learning. Using it as a substitute for cognitive effort is the problem.
The argument I’m trying to make is very distinct. It’s not the child asking, “Can you tell me another way to do this, or another angle to this argument?” The real danger comes from the fact that children might ask, “Solve this for me” or “Write this analysis” and then move on to the next task.
It’s not about assistance. It’s about substitution.
Who gets to think?
Who owns the systems that generate knowledge and who merely consumes what the systems produce?
We have traditionally understood the digital divide as a question of access: who owns a device, connectivity and computing power. But AI could create a more subtle divide.
Imagine two children growing up in the same AI-enabled world. Both have access to ChatGPT or whatever version or new tool that comes up next.
One grows up in an environment where adults insist that she first read the book, formulate her own view and write her own draft.
The other realises AI as a shortcut: why struggle through the book when it can be summarised? Why formulate an argument when the machine can produce one?
Both children have access to the same technology. But they are not developing the same capacities. The first group may use AI to become extraordinarily productive. The second may become extraordinarily dependent. The technology will be the same. But the inequality will be cognitive. The divide will be between those who have learned how to think and those who ask a machine to think for them. And that could become one of the great inequalities of the AI era, if not chosen wisely.
Some of the tech giants in the industry proudly boast how they have been notably cautious about how much technology their own children consume, with limits on screens, phones and social media. Isn’t it suspicious that makers of these technologies are cautious about giving their own children more of it?
Perhaps they understand something that the rest of us are only beginning to confront.
From digital divide to digital colonisation
This is where the conversation needs to become bigger than parenting.
A wealthy child may have excellent teachers, books, stimulating conversations at home, extracurricular activities, tutors, exposure to different cultures and adults who encourage him/her to question everything. AI for them is an additional tool.
But for a child with resource scarcity, AI might be the tool that would actually replace some of those things. This definitely does imply to restrict access to disadvantaged children.
My argument is that if one group of children are taught to think independently and another is increasingly encouraged to outsource thinking to machines, this could be creating a new hierarchy that might be more troubling than a digital divide.
UNESCO has also warned about age-appropriate use and argues that independent generative-AI conversations should have age limits; its guidance stresses human-centred, pedagogically designed use rather than simply putting AI into classrooms.
The metaphor may sound dramatic, but colonialism was not just about controlling resources. Colonialism was also about shaping whose knowledge was valued, what people learned, whose language became authoritative, who was considered capable of producing knowledge rather than just receiving it, while marginalising others.
The concern with AI is not that it is another form of colonialism in the historical sense, but that it could reproduce a similar hierarchy: a small number of actors creating and controlling knowledge systems, while vast populations become dependent on consuming what those systems generate.
There are some struggles that should exist
Some struggles are essential for a child’s development. Children need the opportunity to wrestle with ideas, make mistakes and work things out for themselves.
While I would not advocate for keeping them away from technology, which these days is neither possible nor desirable, the question, instead, is what we choose to outsource to it—and what we must allow children to do for themselves.
Reading deeply.
Writing from scratch.
Forming an argument.
Solving problems.
Getting things wrong.
Changing minds.
And, perhaps most importantly, learning to sit with a question without immediately asking a machine for the answer.
AI can become a child's assistant, critic, tutor and collaborator. One thing it should never become is a substitute brain. The goal should not be maximum efficiency, but maximum development.
What if the biggest inequality AI creates isn't about who has access to it—but who still knows how to think without it?
AI may not democratise intelligence.
It may concentrate it.
Author’s Bio
Atiya works at the intersection of communication, equity, and technology, amplifying underrepresented voices and advancing impact through storytelling, data, and advocacy. Her vision is to build communication ecosystems that do more than inform: they shift power and help build more equitable futures.