A method and system for distributing to a destination a notification of a topic being discussed by two participants of a voice call. Terms are extracted from an audio stream generated during the voice call. A topic is determined based on the extracted terms. A topic notification identifying the topic and at least one of the two participants is sent to a destination. A recipient of the notification may indicate a desire to join the voice call. Upon approval by one of the participants, the recipient may be joined to the voice call.