This study investigated the effectiveness of visible-annotation types used in online discussions to enhance the accuracy of shared knowledge and the level of constructed knowledge in computer-supported collaborative learning environments. Three types of visible-annotation were developed specifically for this study, which had the function of linking learners' annotations to related learning content. The participants were randomly assigned into one of three groups. Each group was provided with different types of visible annotations for building shared knowledge in computer-supported collaborative learning: the TL-type for two content-understanding learning phases; the TLL-type for one concept-understanding learning phase and one content-understanding learning phase; and the C-type for one content-understanding learning phase. After completing the knowledge-sharing stage, the accuracy of the participants' shared knowledge was measured. During the knowledge-construction stage, the participants carried out a lesson-planning task in pairs and constructed knowledge was analyzed based on the lesson plan. The findings revealed that the TLL-type of visible annotation with one concept-understanding learning phase and one contentunderstanding learning phase was the most effective for enhancing the accuracy of shared knowledge and the level of constructed knowledge in the computer-supported collaborative learning environment.