In a realistic context, the online advertisements have constraints such as a certain number of clicks to draw, as well as a lifetime. Furthermore, receiving a click is usually a very rare event. Thus, the problem of choosing which advertisement to display on a web page is inherently dynamic, and intimately combines combinato- rial and statistical issues. We introduce a planning based algorithm for optimizing the display of advertisements and investigate its performance through simulations on a realistic model designed with an important commercial web actor.