84% of students use AI for homework; only 29% of schools have policies

An assistant professor of school psychology at the University of Wisconsin-Stout surveyed 303 educators and school professionals in Wisconsin—including teachers, administrators, IT staff, school psychologists, and counselors—between spring 2025 and spring 2026, along with 132 professionals at schools across the country, to understand how generative AI is affecting K–12 education and student learning. While the results are not nationally representative, they offer a snapshot of how educators are grappling with AI in the classroom.

The survey found that 84% of students use AI for homework. Yet when asked whether their district had a formal AI policy, only 33% of Wisconsin respondents and 29% of national respondents answered yes. Among the concerns educators raised, academic dishonesty and plagiarism topped the list: approximately 65% of Wisconsin respondents and 74% of national respondents identified these as concerns. However, a deeper worry emerged. When asked about difficulty in assessing student learning when AI is used, 47% of Wisconsin respondents and 53% of the national sample selected this as a concern. The problem is concrete: a student can paste a prompt into a chatbot and receive a polished essay, lab summary, or reading response almost instantly. Teachers then face an epistemological puzzle—does the finished work reflect the student's thinking and effort, or what the chatbot generated?

When respondents described the impact of AI on student behavior and mental health, 29% of Wisconsin respondents and 40% of national respondents reported observing "increased student reliance on AI," while 19% and 33%, respectively, selected "reduced critical thinking or problem-solving." Some teachers have turned to AI-detection tools; in a 2025 national survey of sixth- through 12th-grade public school teachers, 43% reported using such apps regularly, while another 27% had tested or experimented with them. However, these tools are unreliable. One study of 14 AI-detection tools found false-positive rates as high as 50% and false-negative rates as high as 100%, depending on the tool. About 20% of AI-generated texts were misclassified as human-written; that rose to about 52% when AI-written text was manually edited and 71% when machine-paraphrased. Other researchers found that detectors falsely flagged nonnative English writing as AI-generated at an average rate of 61.3%.

To address these challenges, some teachers are redesigning assignments. They ask students to show or explain their process, include oral components alongside written work, or write more in class. When the goal is independent thinking, some assign paper-and-pencil tasks. Researchers who developed the Artificial Intelligence Assessment Scale have argued that educators should identify what level of AI use makes sense based on the learning outcomes being measured—recognizing that one assignment might forbid AI entirely to see independent writing, another might allow AI for brainstorming but require original notes and reflection, and yet another might ask students to critique an AI-generated answer. The educators surveyed were not uniformly rejecting AI; many reported using it themselves for planning, communication, documentation, and administrative tasks. Their concerns pointed to a practical challenge: how to preserve meaningful evidence of learning when AI can produce polished academic work in seconds.

The tension at the heart of this survey is not new to education—teachers have always struggled to verify that finished work represents genuine student understanding. A parent's help, copying from peers, or rote completion without comprehension have long made it hard to interpret assignments as direct evidence of learning. Generative AI amplifies this problem by making polished, seemingly original work instant and nearly indistinguishable from human effort. The gap between policy and practice is stark: 84% of students use AI for homework, yet fewer than one in three districts have set formal rules for when and how it is acceptable. This creates ambiguity not only for teachers trying to assess learning but also for students unsure what is permitted.

The survey reveals that educators' worries extend beyond cheating. While 65–74% cited academic dishonesty and plagiarism as concerns, roughly half of respondents flagged a more fundamental problem: the inability to know whether a student understands what they submit. One-third to two-fifths of respondents observed increased reliance on AI and reduced critical thinking among students. These findings suggest that the challenge is not policing misuse but designing learning tasks that still yield reliable evidence of understanding—a shift some teachers are already making by incorporating oral explanations, in-class writing, process documentation, or assignments that require students to critique AI output rather than generate it from scratch.