Purpose: Empirical validation of educational technology is critical for best practice, particularly when courses are delivered online. This study aims to investigate the predictive relationship of usage behaviors on perceptions of 30 preservice special education teachers while reading in an online social annotation reading tool. Design/methodology/approach: In this single-group quasi-experimental study, participants completed two readings in Perusall, once individually and once in small groups, then took a researcher-created survey after each reading. Descriptive data and paired sample t-tests were calculated. Predictive relationships between usage behaviors and survey results were analyzed with linear regression models. Findings: Participants thought Perusall was useful for their learning and easier to use in small groups and that guided reading prompts were helpful. Usage behaviors did not significantly account for participant beliefs about Perusall. Instructors may wish to use guided reading prompts and small groups to maximize student learning and engagement. Originality/value: This study addresses gaps in the literature (Suhre et al., 2019; Sun et al., 2023) by following one group of students over two semesters, using a commercially available tool, measuring actual usage behaviors and not solely student perceptions and analyzing instructor perceptions of the tool. The authors contribute further evidence that group-constructed knowledge is valuable for undergraduate learning (Kalir et al., 2020b). The authors also provide data-based suggestions for the use of social annotation tools that maximize student learning and engagement.