The sharing of accommodation as a sustainable environmental solution for the lodging market is prevalent all over the world. However, the rapid expansion and low occupancy rate could be due to the accommodation hosts’ lack of attention to the pre-interaction content while reducing their carbon footprint, which have caused a significant impact on guests’ decision-making and prevented sharing accommodation. To improve the host pre-interaction capabilities and achieve the environmentally friendly potential of sharing accommodation, this research aims to explore the host expression characteristics including important topics and keywords of host pre-interaction content from a symbolic interaction perspective. Conducting the latent Dirichlet allocation machine learning model, keywords clustering characteristics emerged as main topics based on 38,814 listings from Airbnb in Beijing. The result of investigating the features in these topics and the word distribution in three types of properties shows that in a homogenous accommodation community, hosts who make the pre-interaction have more orders than those who do not. At the same time, the focus of hosts on expressing explicit and abundant topic symbols can effectively increase the attractiveness of their listings. However, accommodation hosts who just post a long text but do not emphasize listing key topics would not convince guests to use the accommodation. A variety of practical implications of findings has been discussed for sharing accommodation practitioners to answer the challenge of sustainability.