Schools need a grade-by-grade strategy for AI that matches how children's brains develop, not a one-size-fits-all approach. A new framework suggests keeping artificial intelligence out of students' hands in kindergarten through second grade, gradually introducing it under teacher supervision in elementary school, and only allowing independent AI use once students reach high school. The key principle: students must do the thinking themselves first, or they won't develop the skills they need.

Why Does AI Pose a Unique Challenge in Schools?

Unlike previous classroom technologies such as computers or interactive whiteboards, generative AI can actually perform the intellectual work that lessons are designed to teach. It can write essays, solve math problems, summarize readings, and produce analyses that students are supposed to complete themselves. This creates a fundamental problem: if a student hands off their thinking to an AI tool, they may never develop the cognitive skills the assignment was meant to build.

Research on what experts call "cognitive offloading" reveals the stakes. In a randomized study involving about 1,000 high school math students, researchers compared two versions of an AI chatbot. One answered questions directly. The other offered hints and guidance but withheld the final answer. Students using the first chatbot performed worse on later exams than students who didn't use AI at all. Students who received only hints performed as well as the no-AI group. The lesson was clear: when a tool does the thinking for students, they learn less.

"The student must do the thinking," explained Jennifer Weber, a K-12 education policy fellow at the Manhattan Institute and author of a forthcoming brief on AI in schools.

Jennifer Weber, K-12 Education Policy Fellow, Manhattan Institute

How Should AI Use Change Across Grade Levels?

The developmental framework suggests a carefully calibrated approach that expands AI's role as students mature and build foundational skills. The progression reflects how children's cognitive abilities develop and what research shows about when different types of learning become appropriate.

• Kindergarten Through Second Grade: AI should remain exclusively in teachers' hands. Young students are building the foundational skills that everything else depends on, including sounding out words, forming letters, and developing number sense. These early literacy and math skills predict academic performance years later. When AI gives students answers, the productive struggle that builds these skills never happens. Teachers can use AI to plan phonics instruction and generate practice problems tailored to each student's level, but students should never use it directly.
• Grades Three Through Five: Students can begin using AI, but only after completing their own work first and only under direct teacher supervision. A teacher might take a paragraph a student has written, generate an AI version, and have the class identify errors, weak reasoning, and unsupported claims. Research shows that when students critique AI-generated work with teacher guidance, their critical thinking improves. This approach leverages AI's ability to produce examples for analysis without replacing student thinking.
• Middle School: AI should become something students study, not just something they use. This is when students should develop the judgment to evaluate information critically. Every eighth grader should understand how AI technologies work, where they fail, and why they can produce convincing but incorrect text. This marks the beginning of true AI literacy. Students' own work should still come first, with AI used afterward and under teacher direction.
• High School: Students who can read critically, write clearly, and work through problems independently can get real value from AI. In writing, they can draft an essay first, then use AI to identify gaps or missing arguments, deciding whether to follow the tool's suggestions. AI can also function as a tutor, helping students catch up in subjects like algebra or progress beyond grade level, providing feedback faster than a teacher managing 25 students. High schoolers are entering a world where AI is embedded in most professional fields, so they need to learn to use it effectively while maintaining the ability to think, write, and reason without it.

What Does the Research Actually Show About AI and Learning?

The framework draws on decades of learning science, including foundational work by psychologist B.F. Skinner, who in 1958 designed a teaching machine based on a simple principle: students couldn't move forward until they produced the correct response themselves. The machine provided immediate feedback but never completed the work for the student. Modern research on cognitive offloading confirms that the more students hand tasks to machines, the less able they become to do those tasks themselves.

A meta-analysis cited in the framework found that when students critique AI-generated work with teacher guidance, their critical thinking improves. The same research points in the opposite direction too: the more readily people trust AI's output without questioning it, the less they think for themselves. This suggests that AI's value in schools depends entirely on how it's used.

Steps to Implement an Age-Appropriate AI Framework in Schools

• Establish Clear Assessment Rules: Keep all assessments AI-free across every grade level. Tests should measure what students can do, not what a chatbot knows. This ensures teachers have accurate information about student learning and prevents AI from inflating grades or masking gaps in understanding.
• Train Teachers on Cognitive Offloading: Educators need to understand the research showing that when AI completes work for students, learning suffers. Teachers should learn to design AI activities that require students to think first, critique AI output, or use AI as feedback after independent work, rather than as a shortcut.
• Develop Grade-Specific Guidelines: Create concrete policies for each grade band that specify when and how AI can be used. More than 30 states have now issued AI guidance for schools, but much of it focuses on data privacy and teacher training rather than developmental appropriateness. Schools should go further and define what thinking students should do themselves at each stage.
• Prioritize AI Literacy in Middle School: Make AI literacy a core part of the curriculum by eighth grade. Students should understand how these systems work, their limitations, and why they sometimes produce convincing but false information. This foundation prepares students to use AI responsibly in high school and beyond.

The overarching principle is simple but powerful: AI's role should expand only as a child's own foundation grows. Technology has never been the goal of education; learning is. Schools exist to help students develop knowledge and skills they couldn't develop on their own, and every technology should earn its place by helping accomplish that goal. AI may become one of the most powerful instructional tools schools have ever adopted, but its success shouldn't be measured by what the technology can do. It should be measured by what students can do as a result of it.