Economy for Everyone
Schools reached for bans and detectors and called that a plan.
Students kept using AI anyway - at roughly the same rate whether the ban was in place or not. We’ve seen enough to say the ban-and-detection approach doesn’t work. It can’t be enforced, and the detection tools used to back it up have documented false positive rates and documented bias against non-native English speakers.
Teachers who tried to enforce it didn’t build the bureaucracy that handed them these rules. They were given a patch, told to maintain the old system, and asked to make it work in a world that had already changed.
What happened
When a student uses AI to write an essay, the essay may look fine. The problem is what didn’t happen: they didn’t build the argument, didn’t hit the wall where the logic breaks, didn’t restructure, didn’t decide what they think. That process is the learning. Skipping it produces a better essay and a less capable student.
Students who used AI to work through problems performed measurably worse when tested on the same material after the tool was removed. The output improved. The learning didn’t happen. Call it the performance paradox. It’s running right now in classrooms that don’t have a clear line about when AI use crosses from helpful to harmful.
One more number worth sitting with: nine percent of parents believe their teenager regularly uses AI for schoolwork. That number is nowhere near 9 percent. It’s closer to two thirds. That’s why so many adults think the policy worked: they can’t see what’s happening.
Why the policy isn’t solving it
States are now requiring districts to publish AI policies. That’s better than nothing; at least it creates something to audit. But read the laws carefully: they require a policy. They don’t fund one. They don’t require the policy to address what students are supposed to learn, or how teachers are supposed to redesign assessment while running on 4.4 hours of planning time per week.
The policy layer is arriving. The support layer is not.
Teachers in low-poverty districts received AI training at nearly twice the rate of teachers in high-poverty districts. The schools that most need to get this right are the least equipped to do it. The kids in those schools aren’t waiting for the policy to catch up. They’re in class right now.
That’s the failure. The world changed fast. The system did what large systems often do: preserve the status quo, add compliance language, and call it a response. Schools don’t need more compliance theater. They need room to change what students are being asked to do.
What helps
“Write a three-page paper on X” is now a one-minute task. “Make an argument, show your first draft, explain what you changed and why, and tell me why the opposing view is wrong” still takes thinking. In math it lands differently: “Show me the step where your first approach broke down. What constraint did you forget to account for?” Different subject, same test: whether the student understands what they did.
Math has an extra layer. When AI solves the problem, students get the answer and skip the productive struggle that builds fluency and judgment. A student may understand an AI-generated solution while reading it and still be unable to solve the next problem alone. That’s like watching someone else lift weights and calling it exercise. For math, the check cannot stop at “explain this answer.” It has to include transfer: change the numbers, change the condition, ask for an error check, or have the student solve a nearby problem without the tool.
Schools owe kids work that requires thinking. The check is simple, but it cannot be shallow: ask the student to explain what they submitted, then apply the same idea one step away. Make it part of the assignment. A short conversation, a few targeted questions, a brief written defense. Timed checks. Short oral defenses. Revision memos. Tool-removal checks. None of this is magic. It’s measuring whether the student knows the material. These aren’t full-system fixes. They’re realistic moves teachers can make inside a system that hasn’t given them enough time, training, or cover. Explanation is the first check. Transfer is the better one. If they can’t do either, the tool helped them avoid the work that was supposed to happen.
What you can do
If you are a teacher:
You’re already being asked to carry too much of this on your own. It’s one practical move inside an impossible situation. Add one question to one upcoming assignment: “Walk me through how you approached this. What did you try first? What was wrong? What did you change? Now apply the same idea to this nearby problem.” You don’t need a new policy to ask that question. The conversation that follows teaches and tests at the same time, and tells you immediately whether the work was theirs.
If you are a parent:
Most parents are behind the reality of the tools. That’s understandable. The shift happened fast. But paying attention here can help your kid more than you think. Ask two questions about their last big assignment: “What were you trying to prove, solve, or explain?” and “What changed after your first attempt?” You’re trying to understand what they’re building. If they can answer, something is working. If they can’t, it’s worth a conversation with their teacher. Come curious: “How is the class handling AI use? What does the school expect students to be able to do on their own?” Enough parents asking that question can move policy.
One steady action to take this week
Pick one assignment this week, yours to give or your child’s to discuss, and add one question: “How did you approach this? What did you try first? What did you change?” You don’t need a policy to ask that. The answer tells you what you need to know.
Action ladder
Short term
Teachers:Add one verification question to one assignment this week: “What did you change?” “What did you try first?” “Can you apply the same idea to a nearby problem?” Start small and make the explanation part of the work.Parents:Ask those same two questions about one recent assignment and listen for whether your kid can explain the thinking behind the finished product.
Medium term
Teachers:In your team, department, or grade level, push for one shared change this semester: a short defense, a revision memo, or a tool-removal check on major assignments so the burden doesn’t sit on one classroom alone.Parents:Ask the school, parent-teacher association (PTA), or principal a concrete question: how is the school verifying student understanding when AI can produce polished work? Push for major assignments to include explanation, revision, or in-class verification.
Long term
Teachers and administrators:Push districts to fund teacher planning time, AI-era assessment redesign, and practical professional development. A policy without time and support is paperwork. Push testing bodies too: if the Advanced Placement (AP) exam and state accountability tests still reward what AI can produce, classroom redesign can’t propagate up the chain.Parents and community members:Back schools when they ask for the capacity to redesign the work. More rules, more detectors, and more pressure on individual teachers won’t fix a broken system.
How to talk about it
You don’t have to be anti-AI to care about this. The question is whether using the tools produced learning or just output.
A useful line: “If a student can’t explain what they submitted, the tool didn’t help them learn. It helped them avoid learning.”
What matters is whether the student can explain the work, apply the idea again, and know when the answer stops making sense.
The schools moving in the right direction are teaching students to use AI as a starting point, then asking them to prove they went further. That’s the goal. Support teachers. Push administrators for room to redesign the work. Push parents to stop pretending the old rules are enough.