If you’ve turned to ChatGPT or Claude to help shape your writing, recent research suggests the sleek results you’re seeing could be inflating your confidence. A study published in the *Computers and Composition* journal identifies an “fluency trap” created by AI writing tools, where polished, confident‑sounding prose hides superficial thinking and gives writers the illusion that the work is complete.
Fluency ≠ Completion
Abram Anders, an associate professor of English at Iowa State University, and co‑author Emily Dux Speltz, an assistant professor in the Department of Humanities and Communication at Embry‑Riddle Aeronautical University, tracked 38 undergraduates over two semesters in an experimental “AI and Writing” course. Students entered the class expecting AI to slash their workload, but the reality was different.
The study explains that the fluency trap emerges because AI generates text that sounds confident and tidy, prompting writers to trust it even when the content is inaccurate, shallow, or off‑target. Many participants initially approached AI like a search engine—feeding vague prompts and accepting whatever was returned. Over time they discovered that effective prompting demands planning, clarity, and rhetorical awareness—the very skills strong writers already employ without AI.
“AI writes in confident sentences, uses the right tone and sounds smart,” Anders said. “But that polish can trick students into trusting it, even when it’s wrong, shallow, or missing the point entirely.”
What Effective AI‑Assisted Writing Looks Like
The researchers pinpointed three prerequisites for using AI wisely. First, collaborating with AI involves genuine trial‑and‑error, not a single prompt and acceptance. Second, AI output still requires human judgment to verify claims, sharpen logic, and align with the expectations of the specific context. Third, while AI can produce text, it cannot generate purpose; only the writer decides the argument and its purpose.
Students who mastered these three steps stopped treating AI as a shortcut and began using it to explore ideas, weigh options, and refine their arguments. Anders and Dux Speltz describe this evolution as shifting from outsourcing writing to orchestrating it.
“AI changes the workflow, but it doesn’t change the fact that writing is thinking,” Anders noted. This distinction grows more crucial as AI‑generated prose becomes increasingly indistinguishable from human‑written work.