A set of user pools can be determined based on location data associated with each device in an audio/video (A/V) conference. A key active user can be determined for each user pool of the set of user pools based on valid audio signals received from each device within each user pool. A determination can be made whether there is feedback within each user pool. Responsive to determining feedback in at least one user pool, speakers of devices within the at least one user pool can be disconnected except for the key active user device within each respective user pool.