CSCL can foster in-depth knowledge acquisition if learners build on each other’s knowledge to draw new conclusions. Individually preparing learners with generative tasks has been proposed to foster such interactive activities in subsequent collaboration. However, the scarce evidence provides no clear picture. Thus, we conducted an experimental study where participants went through the following phases prescribed by a CSCL-script: after reading a text learners prepared individually by answering a task in written form, read the co-learner´s task answer, and subsequently discussed the text via chat. According to experimental condition, the preparation phases consisted of either non-generative, compare-and-contrast, or explanation tasks. Results revealed no overall effect of preparation task type on interactive discussion activities. However, generative preparation tasks affected interactive activities indirectly through the learners’ and their co-learners’ generative preparation activities. This suggests generative tasks can foster interactive discussion as far as the dyad partners actually enact generative preparation activities.