A randomized controlled trial involving several hundred secondary school students found that AI-generated feedback produced significant improvements not just in the technical quality of student revisions, but in learning motivation and positive affect as well, students' actual emotional engagement with the writing and revision process. This is a genuinely notable finding because it moves beyond the question most AI feedback research focuses on, does the writing get objectively better, to a related but distinct question that matters enormously for sustained learning: do students actually feel more motivated and positive about engaging with feedback and revision at all.

This distinction matters because feedback that improves a single piece of writing but leaves a student feeling discouraged or disengaged from the revision process has real limits as a learning tool; students who feel motivated and positively engaged with feedback are considerably more likely to actually apply what they learn to future writing, beyond just the specific assignment the feedback was attached to. A study finding gains in both technical revision quality and motivational engagement offers considerably stronger evidence for AI-assisted feedback's genuine pedagogical value than a study measuring revision quality alone.

The randomized controlled trial design here also matters for how much confidence to place in the finding: unlike observational studies that can only show correlation, a proper RCT, randomly assigning students to receive AI-generated feedback versus a comparison condition, offers considerably stronger evidence that the feedback itself, not some other pre-existing difference between students, produced the observed improvements.

Why the motivation finding may matter more than the revision finding alone

Revision quality improvements on a single assignment are valuable, but they're also, in a sense, the easier outcome to influence, since even generic feedback often produces some revision improvement. Motivation and positive affect are harder outcomes to move, and they're arguably more important for long-term writing development, since a student who stays motivated and positively engaged across a whole semester of feedback and revision cycles is likely to develop considerably more as a writer than one who makes strong gains on a single assignment but disengages from the broader revision process afterward.

• Pay attention to how AI-generated feedback is framed and delivered, since tone and framing likely matter for the motivational effects this study documented
• Track student engagement with revision over time, not just single-assignment quality gains, as a fuller measure of feedback effectiveness
• Consider that well-designed AI feedback may offer motivational benefits worth weighing alongside efficiency and consistency benefits
• Use this kind of controlled research, rather than general impressions, when evaluating whether an AI feedback approach is genuinely working
• Watch for replication and extension of this kind of study across different age groups and subject areas as the research base grows

What this suggests about how AI-generated feedback should be framed

This study's motivational findings suggest that how AI-generated feedback is framed and delivered, not just its technical accuracy, plays a real role in its educational value. Feedback that's specific, encouraging, and clearly actionable appears more likely to produce the kind of positive engagement this study documented than feedback that's technically accurate but delivered in a flat, purely evaluative tone. This is a useful design consideration for any teacher personalizing AI-drafted feedback before it reaches a student, since the personalization pass is exactly where tone and framing get shaped.

For teachers using AI-assisted grading tools, this research offers a specific, evidence-based reason to invest real attention in the personalization step, not just checking for accuracy, but actively shaping tone and framing to support genuine student motivation and engagement with the feedback.

A fuller picture of what good feedback accomplishes

This kind of controlled research offers a genuinely fuller, more useful picture of what makes feedback valuable than accuracy or efficiency measures alone: real gains in both the technical quality of student work and students' motivation to keep engaging with the revision process. That combination, documented here with real experimental rigor, is a meaningful data point for anyone thinking seriously about how to use AI-assisted feedback well.