Does critiquing AI-generated writing help students write better paragraphs, and does it matter whether they write in groups or alone?

This study examined the effects of critiquing an AI-generated descriptive paragraph on students’ writing performance and critique transfer, and whether writing mode (group versus individual) moderates the outcomes. Although prior research has shown AI’s potential to improve grammar, vocabulary, and student motivation, most studies treat AI as a passive resource, focus primarily on individual use, and rarely examine whether students’ critical evaluations of AI texts transfer to their own writing. This study bridges the gap through a quasi-experimental study with 56 first-year English majors in two intact classes. They critiqued an AI-generated paragraph before writing two introductions either individually or in groups. Unadjusted tests suggested higher word-count and some structural scores for groups, but these secondary differences did not remain significant after Holm adjustment and are treated as exploratory. Group writers reported higher post-task motivation (ηp² = .28) and self-efficacy (ηp² = .48). Individual writers showed superior critique transfer, with perfect strength transfer and fewer repeated AI errors, especially in rhetorical features. The study highlights how active AI critique, combined with different writing modes, can support outcomes in descriptive writing instruction, with groups showing advantages in affective factors and a descriptive tendency toward greater productivity, and individuals demonstrating stronger transfer of rhetorical strengths.