Advanced Traveller Information Systems (ATIS) can be used to collect and disseminate dynamic information about travel times on highway links. One of the potential uses of these systems is to manage incidents. The objective of this research is to show under what incident conditions is it relevant to provide real time traffic information to travellers.A model that uses graphical queueing techniques is utilized to define cases when ATIS is beneficial and cases when it is not, and to evaluate its benefits as measured by travel time savings. The model is applied to a simple road network with two parallel bottlenecks. We analyze an off-peak incident scenario where a user optimal strategy is implemented to disseminate information only to vehicles equipped with ATIS. The different cases of queue evolution that can result are described, benefits to guided and unguided travellers and the sensitivity of benefits to relevant parameters are also analyzed.It is found that once equilibrium is reached between alternate routes, the rate of diversion from one to the other has to be decreased to maintain it. The implication is that during equilibrium some guided travellers will be diverted to the alternate route while others will be asked to stay on the route where the incident has occurred. It is also found that as long as the fraction of vehicles equipped with ATIS is below a critical value, pc, then the benefits to a guided traveller are maximum and are not affected by the amount of guided traffic. However, benefits to a guided traveller decline when the fraction of guided traffic becomes larger than pc. The critical value, pc, does not depend on incident parameters but it is a function of capacity of the alternate route and corridor demand. System benefits also increase to a maximum as the fraction of guided traffic approaches pc and become constant when this fraction is larger than pc. Therefore, under user optimal strategy, if the fraction of vehicles equipped with ATIS is equal to pc benefits to guided traffic and to the system are maximized simultaneously. The research need is to develop a methodology which can find practical estimates of pc to be used in large scale simulations of real life networks.