What is striking about the current AI moment in education is how thoroughly the institutions were caught mid-step. ChatGPT became publicly available in late 2022, and by spring of 2023 it was already appearing in student work at schools that had no policy about whether using it was permitted. The technology did not ask for a transition period, and the people who run educational institutions worked out their positions in real time, in the open, in front of students who had already formed their own. Most schools landed somewhere in the middle of a policy spectrum and are still there, updating as they learn what the actual problems turn out to be.
Understanding this is worth the effort before evaluating any specific school's approach, because the available responses all have real limitations. Schools that banned AI outright encountered an enforcement problem with no clean technical solution. AI detection tools exist, but they produce false positives at rates that make them unreliable as instruments of judgment, and they have been documented flagging non-native English speakers disproportionately. Schools that allowed AI without guidance produced students using it in ways that varied wildly, from sophisticated and thoughtful to obviously hollow. The deeper issue is that the question "should students use AI?" is actually a cluster of harder questions, each with different answers depending on the subject, the assignment, the student, and what the assignment was designed to develop.
What AI currently looks like in classrooms is a spectrum rather than a binary. At one end, tools like Khan Academy's Khanmigo act as patient tutors, responding to student questions in ways designed to scaffold understanding rather than replace the work of developing it. Adaptive learning platforms (IXL, DreamBox, and others) have used algorithmic personalization for years and are now being marketed under AI branding regardless of whether the underlying system has changed in any meaningful way. At the other end, a student can open ChatGPT, paste in an essay prompt, and receive something that will pass a casual read without reflecting any of the thinking the assignment was meant to build. Between these poles is a broad range of writing assistance, brainstorming support, research summarization, code generation, and translation, each of which sits in a different ethical position depending on what a given assignment was actually trying to develop.
The academic honesty question, more honestly
The "did students use AI to cheat" framing captures something real but is narrower than the actual problem. A student who generates an essay and submits it as their own has done something most educators would call dishonest. A student who uses AI to brainstorm an argument, drafts the essay themselves, then uses AI to check grammar has done something considerably less clear, and reasonable people genuinely disagree about where the line falls. A student who uses AI to understand a concept they could not grasp from the textbook has done something that looks more like research than cheating. All three are "using AI," and treating them as equivalent misses almost everything relevant.
A more generative question than "is AI allowed" is "what is this assignment trying to develop, and can AI do that part?" If the assignment is designed to build the ability to construct a written argument under pressure, having AI construct it defeats the purpose regardless of what the policy document says. If the assignment is designed to build research synthesis skills, and the student used AI to surface sources before evaluating and incorporating them, the assignment's purpose is mostly intact.
One practical reason this matters: AI-written work has recognizable patterns, and both teachers and students who know what to look for can usually identify it. The tells are not about grammar or vocabulary. The tells are about the absence of friction, specificity, and voice.
Schools are slowly moving toward designing assignments that are harder for AI to substitute. Oral defenses, in-class writing, process documentation, and personal reflections all require students to be present in the work in ways that matter for actual learning.
What this means for families
For families, the most useful orientation is probably not monitoring but conversation. Understanding what tools your child is using, what they understand the rules to be, and what they have actually learned by using those tools gives you considerably more information than checking submitted documents for AI traces. The students who use AI most thoughtlessly are usually not doing so out of laziness or dishonesty alone. They are doing so because nobody helped them think through what they were actually trying to get out of the assignment, or because the assignment itself did not make that clear. The school's job is to clarify the rules; the family's job is to help students think about why the rules exist, which is a harder and more durable lesson.