EdCircuit Staff 
Generative AI has made student agency harder to see. Two students can submit equally polished assignments after doing very different amounts of intellectual work. One may have wrestled with the ideas, while the other handed most of the thinking to a tool.
Schools often approach this problem by asking whether students used AI. That matters when an assignment limits or prohibits its use, but it does not tell educators enough. One student might request another explanation or ask for feedback on an argument. Another might have the tool choose the topic, organize the response, and write the draft.
That leaves teachers with a more useful question: Who remained responsible for the thinking?
Generative AI offers faster access to feedback, language support, and new ways to explore a subject. It also makes it easy to avoid uncertainty and accept an answer without fully understanding it. Whether the technology strengthens or weakens agency depends largely on how the learning is designed around it.
Better work does not always mean more learning
Generative AI is good at improving the appearance of student work. It can correct errors, organize information, strengthen a sentence, and suggest details a student overlooked. Those improvements may be useful, but they do not prove that a student has developed greater understanding.
The OECD Digital Education Outlook 2026 makes this distinction clear. Its review found that generative AI can support learning when guided by sound teaching principles. When students outsource a task, it may improve performance without producing meaningful learning gains.
Consider a student who asks an AI tool to explain a math concept in a different way and then solves a new problem independently. The tool has helped the student return to the thinking. If the student copies the original problem into the tool and transfers the answer, the task has been completed without the same intellectual work.
Both pages may contain a correct solution, but only one shows what the student can do.
This is why schools cannot rely exclusively on the quality of a final product. Teachers also need opportunities to see how students framed the problem, selected information, made decisions, and responded when something did not work.
Where AI can expand student agency
Used with a clear purpose, AI can help students take actions that previously felt out of reach.
A learner who is unsure how to begin can ask for questions that narrow a topic without requesting a completed thesis. A multilingual student can compare explanations in two languages. A student with a disability may use speech, text, or visual tools to communicate more independently. Someone who needs practice can request another example and receive immediate feedback.
Students can also compare several approaches, test an explanation for different audiences, or ask the tool to challenge an early assumption.
The value comes from what students do next. They must decide which suggestion fits the goal, reject weak or inaccurate information, and explain why they selected one direction over another.
Prompting alone is not enough. Students need to recognize confident errors, invented sources, and bias. Sometimes the responsible decision is to stop using the tool and return to a primary source, a classmate, or the teacher.
When support starts replacing the student
The risks surrounding AI are larger than academic dishonesty. A student can follow the rules and disclose every use of AI while still giving away too much of the learning process.
Imagine a research assignment in which AI suggests the topic, forms the question, summarizes the sources, creates the outline, and drafts the response. The student may type several prompts and approve each step, but the tool has made most of the decisions that gave the assignment educational value.
AI can also remove productive struggle before it has a chance to do its work. Students need time to sit with a difficult passage, attempt an imperfect solution, notice a contradiction, and decide what kind of help to seek. Those moments build persistence and judgment.
That does not mean struggle should be prolonged for its own sake. Students need appropriate scaffolds, and those supports will look different depending on their needs. The goal is to help students move forward without making every decision for them.
WestEd has recommended using AI to scaffold student choice rather than script every move. That offers a practical dividing line: Give students enough support to continue doing the work themselves.
Make the thinking visible
AI is forcing schools to reconsider what counts as evidence of learning. A final essay, presentation, design, or solution may still matter, but it can no longer be the only evidence a teacher sees.
The practical response is not to turn every assignment into an investigation of possible cheating. It is to create normal opportunities for students to show their thinking as the work develops.
Teachers might use a short proposal conference, an annotated source, an in-class checkpoint, a comparison between drafts, or a brief oral defense. Students can explain what they changed after feedback, which AI suggestions they rejected, and where they verified information.
Disclosure becomes more useful when it is specific. “I used AI” reveals little. A student could instead identify the tool, explain why it was used, describe what was accepted or rejected, and note how the output was checked. That reflection becomes evidence of both AI literacy and student agency.
Local assignments also make thinking easier to see. A student might use AI to compare solutions to a campus problem but still need to gather evidence, speak with people affected by the issue, and defend a recommendation.
Teachers do not need to add every checkpoint to every task. One or two well-placed moments can provide a clearer picture of what the student understands and who is directing the work.
Four questions for classrooms and AI policies
Schools cannot write a separate rule for every tool or new feature. Lists of permitted and prohibited uses provide clarity, but they can become outdated quickly. A few consistent questions can help students and teachers decide whether an AI use preserves agency:
- Who set the purpose? The student should be able to explain what they were trying to understand, create, or improve before asking AI for help.
- Who made the important decisions? The student should identify the choices that shaped the work, including the question, evidence, method, interpretation, or revision.
- Who evaluated the output? AI-generated information must be checked rather than accepted because it sounds polished or confident.
- Who can explain the result? The student should be able to defend the work, describe its limitations, and apply the learning without relying on the tool.
These questions move the conversation beyond detection and return responsibility to the center.
The TeachAI guidance for schools includes maintaining student and teacher agency among its core principles. It recommends that people remain critical consumers of AI and retain responsibility for academic and organizational decisions. UNESCO’s guidance on generative AI also calls for a human-centered, age-appropriate approach that protects privacy.
District leaders should consider more than whether a product is efficient or accurate. Does it encourage students to form questions? Can they challenge its recommendations? Can teachers override automated suggestions? Do students understand what happens to the information they enter?
Equity also deserves close attention. Students with strong background knowledge may be better positioned to catch a weak or inaccurate response. Other students may be more likely to trust an answer that appears authoritative. Access to AI must be accompanied by access to knowledgeable teachers, explicit instruction, and opportunities to practice judgment.
Preparing students to remain decision-makers
Students will live and work with AI. Preparing them does not require keeping every tool outside the classroom or accepting every use simply because the technology will be part of their future.
Schools have a more demanding responsibility: Teach students to use powerful tools while remaining accountable for their decisions.
AI may help a student see possibilities that were previously difficult to reach. The student must still decide what matters, determine what is accurate, and take responsibility for the result.
The future will include AI. The question for schools is whether students will enter that future prepared to make decisions—or waiting for a system to make those decisions for them.
Subscribe to edCircuit to stay up to date on all of our shows, podcasts, news, and thought leadership articles.
