The crypto-token market has received significant attention in recent years from investors, policymakers, and scholars. Crypto tokens are digital assets that are built on another cryptocurrency’s blockchain and can be used to support the operations of organizations. In this paper, we investigate the relationship between information disseminated on social media and the crypto token market. We collect more than three million tweets and trading activities relating to more than 500 crypto tokens from the first day to six months after they are listed on exchanges. Using the Panel Vector Autoregression model, we find that investor attention, as measured by the volume of tweets, significantly influences the trading volume as well as return of the focal crypto token on the next day. To further understand the mechanisms behind, we use a state-of-the-art language model for Natural Language Processing–sentence BERT to analyze the content of these tweets. Our results suggest that overall, tweets focused on the intrinsic value of the tokens, such as the team, vision, and technology of the projects, have the biggest volume, their volume impacts the trading volume but not return of the token in the next day. While the volumes of tweets that focus on promotion, trading, general support, and speculation all significantly impact both trading volume and return of the focal token at day interval. These results suggest that the community of investors does care about the intrinsic value of the tokens, although the market might be speculative in the short term. This paper is among the first studies to provide empirical insights into this fast-growing novel market, and the findings can inform policymakers, investors, and startup projects.