A simple, but realistic, model of an autonomous emergency brake (AEB) system was studied. Using Matlab, the model was applied to 543 car‐to‐pedestrian and 607 car‐to‐bicyclist real‐world collisions gathered from the highly detailed German In‐Depth Accident Study Pre‐Crash Matrix (GIDAS PCM) and weighted for representativeness. All collisions were to the front of the car. The aim was to investigate how AEB performance was influenced by varying some of the most relevant system parameters. A reference system was predicted to provide very high effectiveness in saving lives and mitigating severe injuries. However, the effectiveness was substantially impaired by imposing restrictions on functionality in darkness and high speeds. Further, effectiveness was highly sensitive to timing of brake activation and deceleration provided by the AEB system. Combining all these restrictions (darkness, high speed, timing and deceleration) led to a tenfold decrease of effectiveness compared to the reference system.