In Japan, CLIL instruction falls under a soft-CLIL approach, content serving as secondary to language instruction. Furthermore, assessment in classrooms in Japan is oftentimes limited to assessing the product summatively. In the paper, we argue for the value of focusing on content in CLIL activities and assessing the process with the goal to promote learning. The present small-scale study at a Japanese university explored how learners (n = 6) used multimodal mediational means to build their conceptual understanding of ‘Earth breathing’ in order to create a presentation on it for a general English course. The further goal was to explore how inferences made from assessing this process of learners co-constructing their understanding can benefit the formative assessment of the outcome of their collaboration. We analysed learners’ face-to-face classroom interaction and forum posts using mediated action as the unit of analysis. The findings revealed that through building on multimodal mediational means, learners were able to build their conceptual understanding and use academic language with this understanding. Deeper insights into learner performance were obtained from assessing the process of their collaboration. We will discuss the implications of the findings for English as a foreign language (EFL) and CLIL classrooms in Japan and beyond.