AI feedback helps students polish their writing quickly, but relying on it across an entire semester triggers diminishing returns. When students stop questioning automated suggestions, their critical reflection stalls and feedback fatigue sets in.
AI feedback speeds up basic writing edits, but its perceived value drops over the semester as students tire of generic advice and switch to autopilot.
Middle school, high school, and college students increasingly treat large language models as 24/7 writing tutors. If your child assumes every critique from ChatGPT is correct, they hand over their editorial judgment to software.
The danger is not just bad advice; it is passive consumption. When kids rely on automated red pens for every assignment, they optimize for completing the task quickly rather than understanding why a revision makes their writing clearer.
Most research on generative AI in education examines a single assignment or a one-off laboratory test. Those snapshots usually show high enthusiasm and quick gains.
The researchers tracked repeated use over several months to answer a critical practical question: What happens to student engagement and reflection when automated coaching becomes an everyday routine?
- Immediate advice proved popular: Across nearly 3,000 feedback instances, students initially rated AI suggestions as actionable and helpful for surface-level edits.
- Dissatisfaction climbed over time: About one in ten feedback sessions was rated unhelpful, with negative evaluations rising steadily toward the end of the term.
- Speed trumped reflection: AI offered fast editorial tips, but it did not automatically push students to reflect more deeply on their arguments.
- Selectivity made the difference: The students who gained real value were those who engaged critically with the machine, treating suggestions as options rather than instructions.
Automated feedback creates an illusion of competence. In September, getting instantaneous suggestions feels like a breakthrough. By November, the repetitive cadence and predictable phrasing of AI critiques wear thin, causing feedback fatigue.
When software provides answers in seconds, students naturally take the path of least resistance. Without explicit instruction to interrogate what the bot says, students stop wrestling with their own sentences—the exact struggle where real writing development happens.
This study analyzed 283 university students in Estonia, not younger K-12 children whose self-regulation and reading comprehension are still maturing.
The paper is also an unreviewed preprint, meaning independent peer experts have not yet formally evaluated the methodology. In addition, students picked their own AI tools, so feedback quality varied across different models and interfaces.
- If your teen uses AI to check an essay draft, prompt them to ask for one narrow critique—like identifying weak transitions—rather than requesting a full-paper edit.
- If your child blindly accepts the tool's phrasing, ask them to point out two suggestions they deliberately rejected and explain why.
- If your student hits a mid-semester writing slump, take AI completely off the table for one assignment to reset their editorial confidence and clear out feedback fatigue.
- If you want to test whether the tool is actually helping, check if your child can explain their thesis in their own words without looking at the screen.
Use AI as a sounding board, not an editor-in-chief. The moment your child stops questioning the tool's suggestions, it ceases to build their writing skills.
Andres Karjus, Janika Leoste, Tiia Õun (2026). Student Evaluation of Repeated AI Feedback Across a Semester of Writing. arXiv (preprint). — arxiv.org



