Ohio’s public school districts, community schools, and STEM schools have met the state law requiring them to adopt an artificial intelligence policy by July 1, 2026. With policies now in place, the focus is turning to how these guidelines will effectively support students and teachers as they return to classrooms.
The state’s model policy, developed by the Department of Education and Workforce, provides a framework for districts. It addresses key areas such as clear rules for student and staff use, privacy, ethical considerations, the use of third-party tools, teacher practices, and how AI impacts learning objectives and assessment. However, a central challenge remains for teachers: determining whether AI genuinely aided a student's learning process or simply assisted in producing an assignment.
To address this, a proposal suggests that Ohio should implement a “proof-of-learning standard” in the next phase of AI policy implementation. This standard would apply when artificial intelligence significantly contributes to a graded assignment. In such cases, students would be required to provide a brief explanation covering four specific points:
1. What they asked the AI system to do. 2. What aspects of the AI-generated output they changed or rejected. 3. What information or conclusions they independently verified. 4. What they can now explain or perform without relying on the AI tool.
This standard is intended to be narrowly applied, not extending to common tools like spellcheck, autocomplete, or routine formatting. Instead, it would be relevant when AI plays a material role in shaping a student’s reasoning, research, writing, code, design, or the conclusions submitted for evaluation. The approach aligns with concerns already identified by Ohio’s model policy, which specifically asks districts to consider AI’s effect on student learning objectives and assessment.
A proof-of-learning note would transform this broad principle into a practical tool for educators, offering a "small window into the student’s judgment." This insight is crucial because generative AI tools have the capacity to make "weak understanding look deceptively strong." A student might receive a fluent answer from an AI system without possessing enough knowledge to identify a flawed premise, a fabricated source, or a shallow explanation.
Experts suggest that focusing mainly on detecting AI use can turn the classroom into a contest over concealment. Instead, the focus should shift to a different question: what intellectual work did the student still do? The four-part note can answer this question without requiring surveillance. For example, a history student might explain that an AI system suggested three causes for an event, but the student rejected one after reviewing the assigned sources. Similarly, a computer science student could note that generated code failed an edge case and describe the fix. A career-technical student might show how an AI-generated procedure was modified after comparing it with a safety standard or equipment manual. The evidence of learning would reside in the student's decisions, rather than in chat logs or screenshots.
While the Ohio Capital Journal reported in May that broader efforts to regulate artificial intelligence in Ohio had stalled due to uncertainty over enforceability, the situation in schools is different. Ohio has already taken action, with the legislature setting the policy deadline, the Department of Education and Workforce producing a model, and districts now holding implementation authority. This context makes education a practical area to establish a workable norm of human accountability while larger AI debates continue.
The proposed standard would also serve to better prepare students for the modern workforce. Employers increasingly expect workers to utilize AI tools, but they continue to need individuals who can identify errors, safeguard confidential information, discern when a task requires human intervention, and assume responsibility for the final outcome. Students who regularly practice documenting these critical decisions will enter the workforce with a more valuable skill than simple prompt fluency; they will learn how to supervise a machine.
To prevent the standard from becoming an excessive paperwork burden, the Department of Education and Workforce could publish a concise set of examples illustrating when a proof-of-learning note is appropriate and when it is unnecessary. Teachers could adapt the four questions to fit their specific subjects. Districts might also consider piloting the approach in a limited number of courses during the fall semester, allowing them to compare student work and teacher feedback before expanding its use.
Schools must also prioritize student privacy. Students should not be required to submit full prompt histories or sensitive personal information to demonstrate responsible AI use. The existing state model already emphasizes the importance of privacy and protecting personally identifiable information. A proof-of-learning note should document human decisions, rather than creating a new archive of student conversations with AI systems.
The July deadline ensured that Ohio schools adopted their AI policies. The upcoming school year will serve as a test of whether these policies effectively enhance learning. By asking students for evidence of their judgment whenever AI plays a significant role in their academic work, Ohio can make these policies more useful. The ultimate goal is clear: students may learn with powerful tools, but they must still be able to demonstrate the thinking that is uniquely their own.
