A method for distributing voice-recognition grammars includes receiving match data from a first remote element. The match data includes information associated with an attempt by the remote element to match received audio information to first stored audio data. The method also includes generating a grammar entry based on the match data. The grammar entry includes second stored audio data and a word identifier associated with the second stored audio data. Additionally, the method includes transmitting the grammar entry to a second remote element.