The Four Ways AI Is Making Your Kid Worse at Thinking (And What to Do About It)

Here’s a number worth sitting with: more than half of school-age children are already using generative AI tools. Only 31% of their parents know about it.

That’s not a statistic from some edge-case population. That’s from an NSPCC report published in January 2025, and it tracks with what teachers are reporting across the country. The tools are free, they’re on every phone and laptop, and kids have been using them for two years. The only question is whether you have any idea what’s going on in your own house — and whether you’ve set any expectations around it.

A peer-reviewed scoping review recently published in Computers and Education: Artificial Intelligence synthesized over 400 studies on how generative AI is affecting K-12 students. The findings aren’t catastrophic — AI isn’t turning kids into zombies. But four specific risks come up consistently enough that every parent with school-age kids should know them. Here’s what the research actually says, translated into language that matters at home.

Your Kid’s Brain Is a Muscle. AI Is Doing the Reps.

The most important concept in this research is one you can explain in a sentence: cognitive debt.

When your kid uses AI to generate an essay, solve a math problem, or figure out how to structure an argument, they’re outsourcing the cognitive work. Not just the output — the thinking itself. And that thinking is how the brain builds the neural pathways for reasoning, memory, and critical analysis.

MIT Media Lab measured this directly. In a 2025 study using EEG brain monitoring, students who habitually relied on ChatGPT showed measurably lower neural activity in the frontal cortex — the region responsible for critical thinking and memory formation — compared to students who did the same tasks without AI assistance. Over four months, the gap widened. The more they used it, the less their brains engaged.

This isn’t a hypothesis or a worried parent’s intuition. It showed up in brain scans.

The analogy that helps: if your kid broke his arm and wore a cast for six months, the muscles in that arm would atrophy. Not because anything was wrong with the arm — because it wasn’t doing the work. Cognitive debt is the same process, playing out in the frontal lobe instead of a bicep. When AI does the reps, the brain doesn’t build the strength.

The Four Risks, Translated

The 400-study review identified four patterns that show up consistently across the research. Here’s what they actually mean for a kid you know.

1. Academic Integrity Erosion

You’ve probably heard about this one. Kids submit AI-generated work as their own. Detection tools are unreliable — teachers mostly catch it when the kid can’t explain what they supposedly wrote.

But the less-discussed side is subtler: even when a kid is technically following the rules, having AI write a first draft that they “edit” means they skipped the part where they figure out how to have an original thought and articulate it. The output exists. The learning didn’t happen.

Fortune reported in February 2026 on a pattern teachers are seeing across the country: students who can produce an essay with AI but can’t explain the argument in that essay when asked face-to-face. They’ve learned how to produce a document. They haven’t learned how to think.

2. Cognitive Debt

Already covered above — but it’s worth repeating that this is the one the research is clearest on, and the one that’s hardest to see coming. It accumulates quietly over months, not dramatically overnight. Your kid’s grades might not change. Their ability to think without assistance might be eroding anyway.

3. Reduced Creativity and Authorship Voice

Researchers are tracking something that doesn’t make headlines but should: adolescents who use AI heavily for writing are losing their individual voice over time. The outputs are competent. They’re also increasingly homogenous.

This matters beyond academics. A kid who never had to struggle to express something original — who always had a polished AI version immediately available — may reach 18 without knowing what they actually think, or how to say it in their own words. The deficit runs deeper than academic performance; it’s about having a self to express in the first place.

The least intuitive risk: AI interaction substituting for peer collaboration. Kids who can ask an AI instead of working through a problem with a classmate increasingly do exactly that. It’s faster, it doesn’t require navigating a relationship, and it doesn’t come with the friction of another person’s perspective.

Interpersonal work — figuring out how to communicate, how to persuade, how to disagree productively — is how kids build social reasoning. A 2026 Brookings yearlong study spanning 500 interviews across 50 countries found that social relationships ranked among the top concerns parents and teachers identified around student AI use. The concern isn’t kids talking to AI instead of texting friends. It’s kids solving problems alone with AI instead of solving them together with people.

Five Questions to Ask Your Kid Right Now

You don’t need to confiscate devices or deliver a lecture. You need to know where your kid actually is. These questions do that.

“Can you explain to me how you got that answer?”

This is the simplest test. A kid who did the thinking can walk you through it. A kid who offloaded it to AI will hesitate, give vague answers, or essentially re-read you what the AI produced. That tells you something.

“What was hard about this assignment?”

Kids who rely heavily on AI often say “nothing.” Not because they’re confident — because they didn’t actually do the hard part. The struggle is where the learning happens. No struggle is a red flag.

“What did you learn from this homework?”

Simple and obvious. But if the answer is blank stares or a shrug, that’s worth a conversation.

“Did you write a first draft yourself before using any tools?”

This tests whether AI is functioning as a crutch or a finishing tool. A kid who drafts independently and then uses AI to check their work is using it well. A kid who opens ChatGPT first and edits the output is skipping the part that matters.

“What would you do if you didn’t have your phone for this?”

The most revealing question. A kid who has a real answer — “I’d think through it, look it up in the textbook, ask a classmate” — is fine. A kid who genuinely doesn’t know has more dependency than either of you may have realized.

A Note for Dads of Middle Schoolers

If you have a kid between 11 and 14, pay particular attention here. The research is consistent: cognitive debt risk is most acute during middle school years. This is the developmental window where independent reasoning is being built.

Before 11, cognitive frameworks are still being laid down in ways that don’t map as cleanly onto the risks described above. After 14, patterns are more established and kids have more metacognitive awareness of what they’re doing. But in the 11-to-14 window, kids are building the neural infrastructure for adult-level reasoning. Heavy AI reliance during this period isn’t just bad homework practice — the MIT research suggests it may interfere with development in ways that matter long after the assignment is turned in.

If you’re going to be intentional with any age group, middle school is it.

Three Rules That Actually Work

You don’t need a complex framework. You need a few rules that make sense to your kid and that you’ll actually enforce.

Rule 1: Draft Before You AI

For any writing or analysis assignment, the rule is simple: produce your own first draft before any AI tools come out. It doesn’t have to be good. It has to be theirs. Then AI can help with polish, checking logic, finding holes — not replacing the original thinking.

This rule is easy to explain and easy to verify. Ask to see the first draft.

Rule 2: Know Before You Submit

If AI was involved in something you’re submitting, you have to be able to explain it — not just “I checked it,” but actually explain the argument, the math, the reasoning. If you can’t explain it, you didn’t learn it, and it doesn’t go in.

This puts the responsibility in the right place. It’s not “did you use AI?” — which is easy to lie about. It’s “can you explain this?” — which they either can or can’t.

Rule 3: Earn the Shortcut

For schoolwork, AI is a tool for checking and improving your own thinking — not replacing it. Once a kid has demonstrated they understand the material (can explain it, can work a similar problem from scratch), using AI to go further is reasonable. Until then, it’s not.

For personal projects — curiosity, creative work, interests outside school — let AI assist freely. The restriction is specifically for structured learning, where the point of the exercise is the learning, not the output. A blanket “no AI” rule sets up an adversarial dynamic and ignores reality. A rule that says “earn it first” is honest about why the restriction exists, which is something teenagers respond to.

The Bigger Picture

A Brookings Institution study published in January 2026 — a yearlong effort with 500+ interviews across 50 countries — concluded that at this stage, the risks of AI to children outweigh the benefits in educational settings. Not forever. Not a case for banning it. But a clear finding that unrestricted access without parental guidance is the wrong default.

UNICEF’s December 2025 global guidance report found that 8 in 10 parents want more guardrails on AI for their children. You’re not overreacting if you’re thinking about this. Most parents are. The problem is that most parents are thinking about it without knowing what’s actually happening — which puts them back in that 31%.

The goal isn’t raising a kid who can’t use AI. It’s raising a kid who can think without it first. Because the gap between those two things — a kid who uses AI as a tool versus a kid who can only function with it — is going to matter more with every passing year.

DadAI helps fathers have the right conversations at the right time. Subscribe to our weekly briefing →