A University of Florida Professor Stopped Fighting AI in His Classroom: A Peer-Reviewed Study Followed
A University of Florida professor redesigned an assignment to integrate AI rather than ban it, leading to a peer-reviewed study that tracks student AI use at the word level using Grammarly Authorship.
Dr. Brian Harfe, a professor at the University of Florida, redesigned an assignment in his large-enrollment course after noticing AI tools could outperform students on essay prompts. His approach became the focus of a peer-reviewed study published in TechTrends in 2026, offering a data-backed framework for institutions grappling with AI in student writing. Traditional methods like surveys and AI detectors were found ineffective, as surveys rely on unreliable self-reporting and detectors provide probabilistic verdicts without granular insights into AI usage.
Harfe’s revised assignment required students to start with a fully AI-generated draft on human cloning and genetic engineering, then revise it to reflect their own perspectives. The final submission had to represent the student’s views, with no penalties or rewards for AI use. To track changes objectively, Harfe used Grammarly Authorship, a tool that records text provenance at the word level, allowing instructors to see how each submission evolved from AI draft to finished product.
The study analyzed 310 essays from students across seven colleges and over 100 majors, revealing that students kept about 76% of the AI-generated draft on average. Students with higher academic standing revised more extensively, and only two of 310 students submitted a 100% AI-generated essay despite explicit permission to do so. The data suggests that assignment design can shift the dynamic from surveillance to meaningful engagement with AI.
The findings challenge concerns that students will fully outsource their thinking to AI, even when given permission to do so. Harfe emphasizes teaching students responsible AI use, stating that tools like Grammarly Authorship facilitate collaboration between faculty and students in understanding AI’s role in writing. The study, co-authored with Aaron Thomas, was published in TechTrends in April 2026.