Virginia Tech Study Unveils Students' AI Writing Strategies and Accountability

A significant study from Virginia Tech, showcased at the 2026 CHI Conference on Human Factors in Computing Systems, delves into the ways students interact with AI tools while writing essays. This comprehensive research, led by associate professor Sang Won Lee, tracked the behaviors of 77 students as they engaged with a specialized online writing environment, capturing every question posed, response received, and text included in their essays.
AI Interactions: Patterns and Preferences
The findings reveal that nearly half of all interactions with the AI involved students requesting it to generate complete essays. This practice, termed "vibe writing," involved students delegating the drafting and revising of content to the AI. Conversely, some learners employed the AI for more limited purposes, such as brainstorming ideas or proofreading existing drafts. This distinction highlights an important factor: students who used AI primarily for planning and revising maintained a sense of ownership over their essays similar to those who did not use AI at all. In stark contrast, those who relied on the AI for full essay generation reported a significantly lower ownership of their work.
The Impact of Accountability on AI Usage
Another key finding offers valuable insights for educators. During a classroom implementation of a system known as SWAG, students were informed that their AI interactions were monitored and visible to their instructor. This transparency resulted in a marked reduction in requests for complete essays, a trend that continued even when submissions were anonymized. This suggests that students internalized the norm of accountability, rather than merely evading detection. The research team is currently developing a version of the system that integrates with learning management platforms, enabling instructors to define acceptable AI usage and access transparency reports.
For both educators and students, the implications of this research are profound. The focus should shift from whether students use AI to how they utilize these tools. By ensuring AI use is visible, establishing clear guidelines — such as using tools that facilitate guided questioning instead of generating entire texts — and rewarding the writing process rather than just the final product, educators can empower students to engage in authentic learning experiences. It’s important to note that while “nearly half of interactions” provides an insightful perspective, it is more specific than claiming “half of all students,” and given the small sample size, these results serve as an indication of trends rather than definitive conclusions.