Animated pedagogical agents (APAs) have the potential to provide one-on-one, just-in-time instruction, guidance or mentoring in classrooms where such individualized human interactions may be infeasible. Much current APA research focuses on a wide range of design variables tested with small samples or in laboratory settings, while overlooking important practical issues relating to large-scale, school-based implementations. The present study provides an early step in addressing this gap by investigating the patterns, affordances and challenges of sustained classroom use of APAs. During a 15-class-period science curriculum, middle school students in the treatment groups (n[subscript Dr C-1] = 191; n[subscript Dr C-2] = 181) had uninterrupted classroom access to one of the two APAs, while control group students (n = 149) completed the same curriculum without APA access. Usage patterns indicate that students accessed the APAs on a fluctuating, "as needed," basis corresponding to the introductory, information-gathering, and synthesis and reporting segments of the curriculum. Survey results revealed no statistically significant difference in student feelings toward the APAs between the two treatments. While treatment students reported that the APAs were unique, reliable, timely resources, interviews indicate little difference between their experiences with the curriculum and those of the control group. Results presented here provide guidance for researchers and practitioners.