An approach is provided to detect a spam message transmitted over a packetized, network-based telephony system. A request to initiate a first voice communication session is received from a requestor. A voice communications session is established with the requestor over the computer network. An audible ringing tone is sent to the requestor over the established voice communications session. While the audible ringing tone is being sent, audio data that is received from the requestor is analyzed. If the analysis detects an audible message in the audio data, then the requestor is identified as a possible spam source. On the other hand, if the analysis fails to detect the audible message in the audio data then the established voice communications session is handled as a non-spam event.