The introduction of autonomous vehicles (AVs) in urban environments where cycling activity is present has raised the need to investigate potential modifications to urban roads, not only from the perspective of AVs but also from the perspective of cyclists. This study aims to investigate the effect of different road design characteristics and varying traffic penetration rates of AVs, using a bicycle simulator study. 50 participants assessed their perceived level of safety, comfort and stress in 11 randomized scenarios. The scenarios involve a design with sharrows and four designs with separated cycling lanes (with two different cycling lane width and two pavement painting options), with 3 AV traffic penetration rates (TPR): 0 %, 50 %, and 100 %. A series of cumulative link mixed models (CLMM) was estimated to analyze the impact of design characteristics and TPRs on cyclist perceptions. The results revealed that the implementation of AVs in shared road scenarios did not improve cyclist safety. On the contrary, it has significantly reduced the perceived level of comfort and has significantly increased the perceived level of stress. However, in separated cycling lane designs, the presence of AVs was found to positively affect cyclist perceptions, although this impact was not significant. Furthermore, the study revealed that the most important factor that affects perceptions of safety, comfort, and stress is the separation between traffic and cycling lanes.