Augmented reality (AR) has been regarded as a useful tool in writing education, with the goal of enhancing students' learning. However, questions still exist about the consistency of student motivation and their writing performance when participating in educational activities driven by AR. This study focused on AR-based writing courses, employing k-means cluster analysis to identify different writing improvement profiles by examining the pre- and post-test scores of 87 primary school students. The study delved into differences in student motivation levels and the frequency of learning behaviors (e.g., disorderly behaviors, raising hands). The analysis identified five different writing improvement profiles, categorized as the Advanced, High-achiever, Persistent, Indifferent, and Diligent groups. Different levels of motivation and learning behavior frequency were observed among these groups. The Advanced, High-achiever, and Diligent groups showed significant improvements in certain dimensions of writing motivation (e.g., curiosity, boredom, competition) in the AR-based writing courses, while such improvements were not observed in the Persistent and Indifferent groups. Additionally, the Indifferent group demonstrated a lack of motivation toward AR-based writing courses, evidenced by increased technical difficulties and disorderly behaviors. Variances were also found in students' sequential behavior patterns among the different groups. These findings shed light on the dynamics of motivation and learning behaviors among students in different writing improvement profiles in AR-based writing courses, offering valuable insights for a more comprehensive understanding of the dynamics at play in such educational settings.