This study aimed to examine the types of reader consideration in student-written argumentative essays and to analyze how metadiscourse is used depending on the type of reader consideration. To this end, methods of processing linguistic information were used to analyze students’ writings, in addition to a survey study. Based on the overall context of this abstract, I assumed that clustering (readers’ stances divided into four groups and reader’s consideration divided into three groups) was conducted from the survey results. Next, metadiscourses were extracted from the students’ writings and classified according to the type of reader consideration. This study is significant because, through an inductive and automated method, it reveals how student writers form assumptions about their readers and make considerations for them when no readers are specified, showing that the writers’ characteristics as categorized in this study are associated with considerable differences in how metadiscourse is used during the actual writing process.