Linguistic research on fictional constructed language (conlang) in video game dominantly employs an approach mimetically designed for fictions and films namely a priori and a posteriori. We argue that a priori and a posteriori are unable to comprehensively analyze the relationship between fictional conlangs with game elements. This research aims at constructing a new taxonomy on fictional conlang design approach that adheres specifically to video game. To do so, first, we analyze 62 game titles with 94 fictional conlangs to prove that a priori and a posteriori display deficiencies in pointing their relationship with video game elements. The next step is constructing a taxonomy based on the integration of Crystal's notion on ludic linguistics, language studies for playful purposes, and Aarseth's textonomy, the study on how texts are accessed. The research findings indicate that a priori and a posteriori fail to classify 29 fictional conlangs into either category and to indicate the relationship between the conlangs with game elements. In response to this failure, we construct a taxonomy consisting of three approaches namely interpretive, explorative, and configurative. These three approaches are able to indicate a structural relationship between fictional game conlangs, game genres, and the immersion level of the players and to patch the weaknesses a priori and a posteriori taxonomy has. Linguists, departing from this new taxonomy, might cooperate with game designers to design comprehensive conlangs, specifically designed for video games, which conform and correspond to the game genres and immersion levels.