Despite the growing popularity of corpus linguistics among researchers in recent decades, a corpus-based approach remains largely unknown to most teachers in primary and secondary schools. Drawing on Shulman's concept of pedagogical content knowledge, this study differentiated between two key terms -- corpus literacy and corpus-based language pedagogy -- and investigated how a group of TESOL teacher trainees developed their corpus literacy and corpus-based language pedagogy in a two-step training scheme. The first step, focussing on using corpus data as a learning tool, was conducted in a physical classroom; the second step, focussing on using corpus data as a teaching tool, took place in a virtual online classroom. A mixed methods design, including surveys, interviews and analyses of lesson plans, was used to collect and analyse the data. The findings revealed that most participants gained a good level of corpus literacy as measured by the self-designed survey. They also obtained a good level of initial competence in using corpus-based language pedagogy, as revealed by the rating of their lesson design and the content analyses of the lessons and interview data. The results support a differentiation between corpus literacy and corpus-based language pedagogy, attesting to the effectiveness of the two-step corpus-based teacher training. The study provides several insights regarding how to scaffold teachers in corpus-based training and teach students with corpus resources to address their vocabulary needs and difficulties. Finally, a few issues are raised regarding what teachers may consider when implementing effective corpus-based teaching in school settings.