There is a vast amount of information flow in social media about bitcoin, which may affect investors' decisions. This article investigates whether tweets may affect returns or trade volume changes of bitcoin and, more importantly, whether some Twitter accounts are more influential than other Twitter accounts. We conduct two separate analyses based first on all Twitter accounts and then on the most influential 50 Twitter accounts, which have been selected as such by Unitedtraders. We use the number of positive, negative and neutral tweets by Valence Aware Dictionary and Sentiment Reasoner (VADER) in a logistic model to analyse if tweets have any valuable information about the change in both return and trade volume of bitcoin. Our results indicate that tweets can be used to predict bitcoin returns. Notably, the most influential accounts are the drivers of returns, but all Twitter accounts simply introduce some noise in volatility. This result indicates that following only these 50 most influential accounts may provide the information needed for investors.