GlobeskoolAI Is Doing Your Kid's Math Homework. Here's What That's Actually Costing Them.
When AI does the hard thinking for kids, homework scores go up and exam scores fall. Here's why the struggle is not the obstacle to learning, it is the learning itself.
Your kid finished their math homework in twenty minutes. It used to take an hour. The answers are correct. You’re half-relieved, half-suspicious, and not sure which feeling to trust.
Trust the suspicious one.
The Study Every Parent Should Know About
Researchers from Stockholm University and the University of Hong Kong tracked 26,800 students across 30 months, watching what happened when AI homework tools became part of their routine. The short version: homework scores went up 18 percent and time spent on assignments dropped 30 percent. Sounds like a productivity win.
Then the exams happened. No AI allowed. Monthly exam scores dropped 20 percent within six months of AI adoption. College entrance exam scores fell 18 to 24 percent. And the effect got worse over time, not better.
The researchers attributed roughly 80 percent of the decline to students who outsourced the problem-solving to AI rather than working through it themselves.
They gave it a name: cognitive surrender. The brain stops wrestling with hard problems because a tool will wrestle for it. The homework looks fine. The thinking quietly goes soft.
A separate 2026 preprint from UC Irvine researcher Sina Rismanchian, analyzing millions of interactions on ALEKS (an online math platform used by over four million students), found the same pattern. After ChatGPT launched, time spent on word problems fell 31 percent among high school students. Proctored exam performance on those same problem types followed it down. Graphing problems, which AI handles less easily, held steady. That detail matters. It means the pattern is not random.
Why Math Specifically Reveals the Problem
Math is a useful canary here because the skill is so legible. Either you can reproduce the reasoning on your own or you cannot. There is no middle ground where you sort of understood it.
When a student uses AI to complete a problem set, what they practice is recognizing that AI can solve this type of problem and submitting the result. That is a real skill. It is just not the skill the homework was designed to build.
The friction in math is not an unfortunate design flaw. It is the entire mechanism. When a student gets stuck, tries something wrong, figures out why it was wrong, and adjusts, that sequence is where the learning lives. Remove it and you remove the learning. The grade stays. The skill does not form.
This applies beyond math. Writing, problem-solving, argument construction, any domain where kids are supposed to be building judgment through practice, the same dynamic holds.
The Difference Between Using AI and Surrendering to It
None of this means AI is the enemy. That framing is too easy and also wrong. AI is going to be part of your kid’s working life in a serious way. The question is whether they grow up directing it or depending on it.
Those are genuinely different outcomes. A kid who uses AI as a thinking partner, who checks their own reasoning against it, who uses it to go deeper on something they already understand, that kid is building a real skill. A kid who hands the problem to AI and copies the output is practicing something closer to learned helplessness.
The difference is not about the tool. It is about who is in charge of the thinking.
This is harder to teach than it sounds, because the path of least resistance is always surrender. AI makes surrender fast, clean, and grade-compatible. Schools are not set up to detect it consistently. And kids, being rational humans, will generally take the easier path unless someone has given them a reason to want the harder one.
What Parents Are Actually Navigating Right Now
If your kid is using AI for homework and the work looks great, you are not in a position to assume learning is happening. The homework is a lagging indicator. The exam is the real signal, and by the time that signal arrives, months of habit have already formed.
The uncomfortable reframe: a kid who used AI to produce a correct answer they cannot explain is not much further ahead than a kid who skipped the problem. Both end up at the exam without the skill.
That is a hard thing to sit with if the homework has been looking good.
The practical question is not whether to ban AI (that ship has sailed and the ban would not hold). The practical question is whether your kid is building the judgment to use it well. That requires a different kind of education than most schools are currently designed to deliver.
What Globeskool Is Built Around
Globeskool teaches kids aged 8 to 16 not just core subjects, but the skills schools mostly skip: critical thinking, problem-solving, working with technology as a tool rather than a crutch. The design is built around real projects where the student has to do the actual thinking, make judgment calls, and work through friction rather than around it.
The goal is not to make learning easier. The goal is to make kids better at doing hard things. Those are opposite goals, and most of the current conversation about AI in education is quietly confusing them.
If you are wondering whether your kid is learning to think or learning to avoid thinking, that question is worth five minutes.