We present a new adaptive broadcast dissemination model to support flexible responses to client requests. Several features distinguish our model. First, client queries do not target individual documents, but specify the required information by attributes. Second, clients are satisfied by responses that are sufficiently close to the desired information. Finally, the server in our model solicits randomized feedback from clients to adapt its broadcast program to client needs. Our simulation results show that our model captures the interest patterns of clients more efficiently and more accurately and scales very well with the number of clients, while reducing overall client average waiting times.