AI Doesn’t Remove Writing Skill. For Struggling Writers, It Removes the Block.

What the researchers found in 38 students’ own words

Abram D. Anders, an associate professor of English and professor of innovation at Iowa State University, and Emily Dux Speltz, an assistant professor at Embry-Riddle Aeronautical University, designed an elective ‘AI and Writing’ course to track what actually happens when undergraduates engage seriously with AI tools over a full semester. They followed 38 students from 22 different majors across two semesters, with students completing structured assignments, keeping reflective records, and documenting how their thinking about writing changed as they worked with generative AI.

Most students arrived expecting AI to be faster. The course showed them it demanded more planning and evaluation, not less. The researchers identified three concepts students had to genuinely understand before AI became productive: that AI writing is experimental and requires iteration, that it demands genuine subject knowledge from the student because AI produces fluent-sounding text even when it is wrong, and that the writer must stay in charge of meaning and purpose throughout. The study was published in May 2026 in Computers and Composition, Volume 81.

The bottleneck in writing is not always in the thinking

The Iowa State finding has a straightforward implication for typical learners: AI raises the cognitive stakes rather than lowering them, and students who treat it as a shortcut get less from it than students who stay actively in the driver’s seat. But for a parent whose child struggles with the physical or orthographic act of getting words onto a page, the finding points somewhere more urgent. Writing draws on several systems at once: the ideas and knowledge a writer brings, the language to express them, and the motor or orthographic output system that transcribes thought into text. For many children who struggle with writing, the bottleneck is in that last layer — not in what they know, not in what they think, but in the laborious, friction-filled process of physical or mechanical transcription.

When AI handles the transcription layer, what remains is precisely the work the Iowa State research identifies as the core of productive AI writing: planning what to say, evaluating whether the output reflects your meaning, and revising until it does. Those are not tasks that bypass writing skill. They are the tasks that writing instruction has always prioritized — and they are tasks where many children who struggle with transcription have real, available capacity. The challenge is that schools rarely distinguish between students whose difficulty lies in the output mechanism and students whose difficulty lies in content or thinking. A blanket AI ban in writing keeps the barrier in place for the child who needs it lifted most, while providing little real protection against the students it intends to discourage. A survey from the 2024-25 school year found that 23 percent of teachers reported AI was typically banned at their school — even as nationwide data show usage rates in ban and non-ban schools are virtually indistinguishable.

The enemy here is not AI, and it is not any individual teacher or administrator making policy with good intentions. It is the assumption that ‘writing is hard’ means the same thing for every brain, and that removing any writing friction is equivalent to removing the development of writing skill. For a child whose bottleneck is the motor or orthographic system, removing that friction is not a bypass — it is an accommodation doing exactly what the research says accommodations should do: remove the barrier to the skill, not the requirement to build it.

The question to bring to your child’s IEP or 504 team

The Iowa State study adds an important nuance that parents can take directly into any accommodation conversation. The students who gained the most from AI were those who stayed in charge of the meaning — directing the AI, evaluating its output, and making deliberate choices about what to keep and what to change. That is the goal, not handing off the thinking. For a child whose challenge is in getting words out physically or orthographically, that framing protects the whole purpose of the accommodation: the child remains the author; AI becomes a transcription tool rather than a ghost writer.

Two questions worth bringing to your child’s school: First, does the AI policy distinguish between students who struggle with the motor or orthographic aspects of writing and students who struggle with content or higher-order thinking? Second, if AI is restricted for writing assignments, what accommodation exists for children whose specific challenge is in the transcription layer? The tradeoffs are real on both sides: an AI policy calibrated to every student uniformly protects the learning of some while blocking the access of others. The research gives parents the frame to ask which category their child falls into — and to ask the school to answer specifically.