Understanding the impact of central bank communication on the effective transmission and predictability of monetary policy is paramount. In this paper, we analyze the role of Banco de México’s qualitative communication and its potential scope for guidance by implementing Natural Language Processing. Using a Latent Dirichlet allocation model and dictionary-based sentiment analysis, we develop a Hawkish-Dovish Tone index to measure the bias of Banco de México’s monetary policy statements. We then incorporate this index into an ordinal logit regression to further evaluate the predictive power of Banco de México’s talk. Our findings show that a rate hike (cut) and a hawkish (dovish) message are associated with higher odds of observing a restrictive (accommodative) decision at the next monetary policy meeting, and communication captures information not reflected in traditional macroeconomic variables. Yet, when it comes to accurately anticipate future interest rate decisions, the information provided in policy statements has limited predictive value.