A Trajectory Computation Infrastructure (TCI) contains three main modules: the Aircraft Performance Model (APM) that provides the aircraft performances (thrust, drag, and fuel consumption, among others), the Weather Model (WM) that provides the wind and atmospheric properties, and the Trajectory Engine (TE) that integrates the equations of motion to obtain the predicted trajectory. As part of its Advanced Trajectory Technologies (ATT) activities, Boeing Research & Technology Europe (BR&TE) has developed a TCI that employs the Aircraft Intent Description Language (AIDL) as the main input to the TCI. AIDL is a univocal, rigorous, and standardized method to express aircraft intent (AI), unambiguously determining the desired trajectory. AIDL is based on the simultaneous use of various instructions, each of them closing a single degree of freedom of the aircraft motion. This paper describes the elements involved in the trajectory computation process. It then explains the different ways of modeling a descent trajectory by means of AI. Its main objective is to obtain guidelines on the consequences that differences between the actual and expected inputs to the trajectory computation process have on the resulting predicted trajectories, and how these can vary depending on the choices taken to model the AI. The different AIs result in the same trajectory when combined with the expected weather (wind and atmospheric temperature) and initial conditions (aircraft mass), but the results diverge when confronted with actual weather and initial mass that differ from the expected ones. This paper describes the reasons for this divergence, analyzes the differences in geometry and speed among the resulting trajectories, and explains why some AI options may be more robust than others when modeling descents, and the risks the modeler incurs when employing each of them. Finally, the AIs employed by the Flight Management System (FMS) vertical navigation (VNAV) modes are described. © 2010 IEEE.