Motivational interviewing (MI) is applied in a variety of clinical and coaching models to promote behavior change, with increasing interest in its potential to optimize school-based implementation fidelity. Yet there has been less consideration of fidelity indicators for MI-embedded coaching and links to outcomes. We leveraged secondary data from 151 teachers across 18 schools, which were part of a larger 39 middle school randomized controlled trial of a teacher coaching model, to explore profiles of fidelity and the associations between fidelity and outcomes. We conducted latent profile analysis (LPA) to examine profiles of four components of fidelity (i.e., adherence, dosage, quality, and teachers' responsiveness). Next, we examined whether observed teacher practices and student behaviors varied across fidelity profiles. Because coaches and independent coders reported adherence, we also examined the reliability of retrospective coach adherence ratings. Results indicated that coaches show promise as a reliable rater of adherence. We identify concrete areas to ensure that reliability can be achieved in other contexts. The LPA indicated that there were two (high and lower) fidelity profiles. Statistically significantly fewer instances of student non-cooperation were observed in classrooms where the teacher was engaged in high fidelity coaching, reflecting a large effect size. Moderate-sized, but non-statistically significant, effects also emerged for teacher opportunities to respond and reactive behavior management. Future directions are considered regarding fidelity measurement and how to optimize coaching.