Students expected AI to make writing easy - they were wrong

A new study shows that AI tools do not automatically lighten the burden of writing; instead, they shift the workload onto students who must still shape thinking, make judgments, and revise. Led by Abram Anders and Emily Dux Speltz, the two-semester project followed 38 undergraduates from 22 majors in a course called AI and Writing, documenting how assumptions about writing changed as students used generative tools. The research identifies three threshold concepts: recognizing that AI-written work is inherently experimental; understanding that AI can sound competent without being accurate; and learning to read AI output critically to avoid the fluency trap where polished prose masks weak understanding. Students who embraced critical evaluation progressed, while those who sought a shortcut found the process more demanding, not easier.