Researchers have a more precise map of how college students use AI writing tools — and it suggests the standard ways schools measure that use are broken.
A mixed-methods study at a public minority-serving R1 university surveyed 382 undergraduates and conducted 14 interviews to identify four reliance types: Strategic (34.3%), Instrumental (30.9%), Dialogic (30.4%), and Dependent (4.5%). Strategic users engaged AI most deliberately, treating it as a thinking partner rather than a ghostwriter. Dependent users — the group most schools worry about — made up fewer than 5% of the sample. The research drew on the AI Literacy Framework, Expectancy-Value Theory, and Biggs's Presage-Process-Product model to build and confirm the typology.
The sharper finding is what the data reveals about assessment. Strategic users — the most intellectually independent group — scored lowest on standard outcome measures, because those instruments track how much AI contributed, not how well the student wrote. In other words, schools are inadvertently grading AI productivity, not student thinking. The study also surfaces a group that existing frameworks miss entirely: roughly 13% of students who declined to use AI on ethical grounds.
For context, most prior research on student AI use has measured frequency alone — a blunt instrument that can't distinguish a student who pastes prompts wholesale from one who uses AI to pressure-test their own arguments. That this gap persisted until 2026 says something about how fast the tools outran the pedagogy.