Multimodal text is becoming the main means of constructing meaning in our lives. As the literacy environment of children and adolescents is very different from the environment in the growth process of the older generation, it is very necessary to understand and analyze how they construct the meaning of multimodal text. Therefore, this study proposes using eye-tracking as a research method that can infer the audience’s visual mode perception and cognitive process. This paper presents an experiment using a webcam eye-tracking program as an example, categorizes the format of the audience’s video reading path, and reveals that the audience organizes the video internally and structures the contents.