GlobeskoolAI Is Saving Teachers Time. But Who Is Actually Teaching the Kids?
A University of Pennsylvania randomized trial found that when teachers used AI to generate lesson plans and exams, their students felt less motivated and rated class as less important. Here is what that means for your child.
There is a version of the AI-in-schools story that sounds like good news for everyone. Teachers save hours. Lesson plans get generated in seconds. Feedback scales. Administrators point to efficiency gains. Parents hear that technology is being embraced.
Then a real study comes out, and the picture gets more complicated.
What the Research Actually Found
A team at the University of Pennsylvania ran one of the first rigorous randomized controlled trials of AI tools in real classrooms. The study followed 193 teachers and more than 2,800 middle and high school students over ten weeks. Half the teachers got access to a ChatGPT-based assistant built around their curriculum. The other half taught as normal.
The AI-assisted teachers primarily used the tool to generate lecture notes, assignments, and exams. They saved time. That part worked.
Their students, however, reported that class felt less enjoyable, less interesting, and less important than what students in the control group described. For students whose teachers were already weaker performers before the trial, the effect was sharper. Those students also scored lower on standardized final exams.
Lead researcher Alp Sungu, a professor at Wharton, put it plainly: teachers, like students or coders, might be using AI as a crutch. The tool gets used. The thinking gets skipped.
The Hechinger Report covered the study. It did not get nearly enough attention.
The Distinction That Keeps Getting Skipped
Efficiency for the adult in the room is not the same thing as learning for the child in the seat. Those two outcomes can travel in opposite directions, and apparently sometimes do.
This is not an argument against AI. It is an argument against a very specific mistake: optimizing the teacher’s workflow and calling it an education upgrade.
When the lesson plan is generated, the feedback is automated, and the exam is AI-drafted, something changes in the classroom dynamic. The content arrives pre-packaged. The intellectual work that a teacher does when they figure out how to explain something, anticipate confusion, and adjust in real time, that work is the part that tends to reach students. Strip it out and what remains is a delivery mechanism that happens to have a person standing near it.
Kids notice. Maybe not consciously. But the study suggests they feel it.
The Specific Skills That Cannot Be Automated From the Student’s Side
Here is what makes this more than a story about one study. The skills that will actually matter for the next decade and beyond are exactly the ones that require active intellectual engagement: framing a problem, questioning an output, deciding when to trust a tool and when to push back on it, communicating reasoning to another person.
These are not skills you absorb passively. They do not transfer through a worksheet, AI-generated or otherwise. They develop when a student has to think out loud, make a judgment call, defend a position, or figure out why something went wrong.
A generated rubric is not mentorship. Automated feedback is not a conversation. And a lesson plan that came from a prompt is not a teacher deciding, based on knowing these particular kids, what they need today.
The irony is this: the very moment schools are most flooded with AI-produced content is the moment when kids most need to learn how to think alongside AI rather than simply receive its outputs. Those two things are opposites.
What Parents Should Actually Be Asking
The question is not whether AI is in your child’s classroom. At this point, in most US schools, it almost certainly is in some form. The question worth asking is whether anyone is still doing the teaching, or whether the tool has quietly become the teacher and nobody formally made that decision.
A few things worth noticing at home:
- Is your child describing class as feeling more repetitive or less interesting than before, without a clear reason?
- Does feedback on their work feel generic, like it could apply to anyone?
- Are they being asked to think and respond, or mostly to consume and recall?
None of those observations is a diagnosis. But they are worth paying attention to, because the shift described in this study is not dramatic. It is gradual. It is the slow replacement of intellectual engagement with procedural efficiency, and kids may not have the words for it even when they feel it.
The Part Globeskool Is Built Around
Globeskool exists for kids aged 8 to 16, and the whole premise is that the student has to be in charge of the thinking. Not the passive recipient of content someone (or something) else prepared. The student as the active agent: using tools, yes, but using them with judgment rather than just deferring to their output.
Core subjects matter. But so does the ability to frame a problem, evaluate a source, communicate an idea clearly, and decide when a tool is helping versus when it is doing the thinking for you. Those skills are not extras. They are the point.
AI is not the enemy of good education. A kid who knows how to think with AI, direct it, critique it, and build on it is genuinely better positioned than one who does not. But that kid needs someone in the loop who is actually teaching, not just prompting.
If you want to see where your child stands across core subjects and future skills, the Globeskool assessment takes about five minutes and gives you something concrete to work with.