Concerns surrounding students utilizing artificial intelligence for academic dishonesty have gained significant traction—and with good reason.
With a few simple commands in a chatbot, students can generate anything from a well-crafted paragraph to a comprehensive essay or a detailed summary of their lab work almost instantaneously. This raises a critical question for educators: Is the work truly reflective of a student’s understanding and effort, or is it merely the product of an AI tool?
Recent data from the College Board, which oversees the SAT and AP examinations, indicates that around 84% of high school students reported employing generative AI to assist with their school assignments in 2025.
As an assistant professor specializing in school psychology and the integration of artificial intelligence in K–12 education, I ponder not only the incidence of cheating but also whether actual learning is taking place.
An ongoing concern about academic integrity and plagiarism
To delve deeper into these issues, I recently conducted a survey of public school educators and administrators regarding the impact of generative AI on educational practices, focusing on how it affects student learning.
Spanning from spring 2025 to spring 2026, my research involved 303 education professionals in Wisconsin—this included teachers, administrators, IT staff, and counselors—as well as 132 additional educators from various institutions across the country.
While the findings do not represent a national consensus, they provide valuable insight into the perceptions of K–12 professionals regarding AI and education.
A significant portion of respondents raised issues related to AI, such as biases and misinformation. However, the predominant concerns revolved around academic dishonesty and plagiarism. In Wisconsin, about 65% of those surveyed expressed anxiety over these matters, slightly lower than the 74% recorded nationally.
Furthermore, respondents highlighted a more complex dilemma: How can educators accurately assess a student's comprehension when AI can effortlessly create essays, summaries, or problem-solving steps?
Approximately 47% of Wisconsin respondents identified the challenge of “assessing student learning in light of AI” as a major concern, with this number rising to 53% in the national sample.
Inquiring about AI’s impact on student behavior, mental health, or engagement, 29% of Wisconsin educators and 40% of national respondents indicated an increase in students' reliance on AI. Additionally, 19% and 33%, respectively, noted a decline in critical thinking skills and problem-solving abilities.
Understanding the completed assignments is becoming increasingly difficult
Teachers traditionally understood that a finished assignment doesn't always equate to evidence of genuine learning. Students might receive excessive parental help, copy from peers, or submit work they don’t fully comprehend.
The rise of generative AI amplifies this challenge, complicating how educators evaluate students’ understanding.
For instance, when tasked with writing a paragraph that explains the theme of a short story, students’ submissions might appear thorough and coherent. However, it’s increasingly challenging for teachers to determine if a student actually grasped the story's theme or simply fed a prompt into an AI system.
Some educators are turning to AI-detection software to discern the originality of students' submissions. A national survey in 2025 revealed that 43% of middle and high school teachers regularly utilized such applications, while an additional 27% had trialed or experimented with them.
Yet, these detection tools can yield inaccurate results—one study highlighted that false-positive rates could reach as high as 50%, with false-negative rates soaring to 100%, based upon the specific tool used. Around 20% of AI-generated texts were misclassified as human-created, rising to 52% when AI-generated text was edited and 71% when paraphrased. Additionally, other research indicated that nonnative English writing was falsely flagged as AI-generated at an average rate of 61.3%.
This doesn’t mean that educational institutions should completely eliminate writing assignments or homework. Instead, teachers might need to more deliberately clarify the objectives of each task.
Many educators are already implementing changes to address these challenges, such as asking students to demonstrate or explain their thought processes or integrating oral presentations into their written work.
Some are reverting to traditional paper-and-pencil tasks to better gauge independent student thinking.
If the objective is to enhance writing fluency, teachers may require observable written work; for reading comprehension, students may need to articulate, apply, or defend their reasoning.
Establishing clearer guidelines for assignments could yield benefits
Many school districts are still developing their AI policies. My survey revealed that only 33% of respondents in Wisconsin and 29% nationally indicated their district had a formal policy regarding AI usage.
Providing clearer guidance about when and how students can use AI could greatly benefit both educators and students.
The creators of the Artificial Intelligence Assessment Scale advocate that educators should clarify appropriate levels of AI use based on specific learning outcomes.
This structured approach is valuable because different assignments demand varied levels of AI interaction. Some may require strict originality, while others might permit AI for brainstorming, necessitating the submission of original notes and a final reflection. Alternatively, students might be asked to evaluate an AI-generated response, identifying what aspects are accurate, incomplete, or misleading.
A shift in perspective
The educators surveyed were not outright dismissing AI; rather, many reported incorporating it into their own workflows for planning, communication, documentation, and other tasks.
Their concerns focused more on specific issues, including cheating, assessment accuracy, student dependency on AI, critical thinking abilities, misinformation, and privacy. These challenges underscore the pressing need for educational institutions to create effective strategies that maintain meaningful evidence of student learning in an age where AI can easily produce polished academic work.
Ultimately, the aim isn’t to identify every instance of AI misuse; that would be an unrealistic pursuit. Instead, the focus should be on creating learning experiences that allow educators to address the key question that remains: What genuinely comprehends this student?
