Group consensus decision making problems with multi-granularity linguistic preference information are studied. From two aspects of individual and group, implied importance degree information of experts under preference information is fully exploited, and determined by an optimization model which is constructed based on the individual consistent degree and the similarity degree between individuals and group. An adaptive linguistic consensus model is presented, where non- consensus preferences identification and correction process are guided by experts’ importance degree. A group decision making method which is a guarantee of reaching a certain degree of group consensus before aggregating expert opinions is developed. Finally, a numerical example illustrates the feasibility and effectiveness of the proposed method.