AI is making it much easier for students to produce polished work without doing much learning.
That is a problem. It is also clarifying something schools should have confronted a long time ago: the product was never the point.
New York City's new AI policy draws a line: elementary and middle school students face broad restrictions, while high school students will be allowed more limited use.
This distinction makes sense. Mostly.
For me, a more ideal scenario would allow middle school students to begin developing AI fluency in a controlled space with clear guardrails. At LightHouse, we have recently built a custom platform for exactly this purpose.
The dangers NYC is trying to protect against are real. Young people need time to develop foundational skills, sustained attention, independent thought, creativity, social connection, frustration tolerance, and confidence in their own ability to solve problems.
If AI steps in too early or too often, it can short-circuit exactly the struggle through which those capacities are built. A 2023 study, “Impact of AI Assistance on Student Agency,” of more than 1,600 students across 10 courses found that students tended to rely on AI-assisted feedback rather than learn from it, and when the AI assistance was removed, the knowledge didn’t hold.
As it turns out, productive struggle is a necessary feature of deep learning.
But a blanket ban creates a different problem.
AI is rapidly becoming part of how people write, research, design, code, analyze, create, and communicate. Students need opportunities to develop fluency with these tools before they enter adulthood.
The challenge is to help them use AI without letting it bypass the work that is necessary to building true understanding and real skills.
Young people (and all people!) need to understand what AI does well and where it fails.
We need to know how to question its outputs, recognize hallucinations and bias, verify information, protect privacy, and remember that a confident answer can still be wrong.
But AI fluency is not just knowing how to use the tool. It is also knowing when not to.
The question worth asking is:
What are we actually asking the student to learn?
If the goal is to learn how to construct an argument, having AI construct the argument defeats the point.
If the goal is to develop as a writer, to find language, structure ideas, revise, and make meaning, then some of the struggle is the learning.
Other times, using AI may be completely appropriate. A student might use it to compare approaches, troubleshoot, translate, generate possibilities, interrogate an idea, or extend an ambitious project.
The issue is not simply whether AI touched the work. The issue is what intellectual work the student actually did.
AI fluency means learning how to use a powerful tool without handing over judgment, curiosity, creativity, or responsibility. It means knowing when AI supports the learning, and when it replaces the very thing you are trying to learn.
You can watch someone else do sit-ups all day. Your own abs will remain entirely unimpressed.
As AI makes finished products easier to generate, we need to become more interested in the purpose and personal investment behind the work.
Instead of simply asking whether a student completed the assignment, we can ask:
What are you trying to understand?
What choices did you make?
What can you do now that you could not do before?
Those questions tell us much more about whether learning actually happened.
Because the real measure of learning is not the product AI helped produce.
It is what the student can now do, understand, or create without it.
At LightHouse, we see it constantly: people work harder, think more deeply, and persist longer when they want to, when the work has real meaning for them.
Meaningful work is often the hardest kind.
Learning to write something powerful, understand a problem, build something, produce a live show, conduct research, repair a bike, master a technical skill, or create something that did not exist before can be deeply demanding.
The difference is that there is a reason to do the difficult thing.
There is a question the learner cares about. A skill they want. A problem they are trying to solve. Something they want to make. Someone they want to become.
That matters enormously in the age of AI.
None of this is new. Work that does not matter to the student has never produced much real learning. People have always found ways around it by copying, cramming and forgetting, or simply doing the minimum to get by.
What is different now is how easy the workaround has become.
A student who once had to at least go through the motions can now produce a finished, polished product in seconds, with none of the thinking behind it.
AI has made phoning in meaningless work fast, convincing, and nearly effortless, and this is exposing something important.
AI did not create the problem of meaningless schoolwork. It made the problem impossible to ignore.
If the product can be generated in seconds, what was the learning actually for?
Banning AI for younger children makes sense to me in many contexts. It is a powerful and potentially dangerous tool, including in its capacity to flatten creativity, replace struggle, and interrupt the development of independent thinking.
But at some point, young people have to learn to work with it, or risk being used by it. We owe students the chance to learn how to use AI well and safely.
We also owe them something harder: work worth doing, even when a shortcut is available.
What excites me is that AI is forcing us to stop confusing the production of assignments with learning. We have to ask harder questions and create space for young people to struggle with their answers:
What do you care about?
What do you want to understand, solve, make, or become able to do?
What is getting in your way, and what kind of help would actually help?
Those are the questions LightHouse has been built around from the beginning.
We believe school should help young people become more capable, more curious, more connected, and more able to do meaningful work in the world.
AI does not change that mission. It makes it more urgent.