Central bank communication to the public is important. Not only do central banks need to influence wage and price-setters expectations to fulfill their objectives, the power that central banks wield creates a democratic obligation to communicate to the public. Despite its importance, communication to the general public is far less studied than communication to financial markets. This paper posits that a key channel through which the general public receives central bank communication is through the print media. We examine which features of central bank text are associated with increased reporting in the news. We write down a model of news production and consumption in which news generation is endogenous. We use our model to show that standard econometric techniques will likely (i) provide biased estimates and (ii) fail to deal with the high-dimensionality of the estimation problem. We use computational linguistics to measure the variables in our model for the case of the Bank of England, and double machine learning to estimate the model. We find that not only does the content of the what the Bank of England produce matter for its news coverage, but also the way in which the Bank of England says it.