Disclosed herein are systems, methods, and non-transitory computer-readable storage media for selectively transcribing messages. Five general approaches are disclosed herein. The first approach is directed to checking for a transcription capable client, which transcribes messages when a client device is capable of receiving transcriptions. The second and third approaches are platform-controlled and user-controlled predefined selective transcription. One aspect of this approach is driven by transcription rules. The fourth approach is user-controlled on-demand selective transcription before the message is stored or deposited for transcription. An example of this is a user transferring an incoming caller to voicemail and indicating that the voicemail be transcribed. The fifth approach is user-controlled on-demand selective transcription after the message is stored. In one embodiment of this approach, a user must specifically request that a stored message be transcribed.