Automation transparency provides an effective way to foster user trust in autonomous vehicles (AVs). However, it remains unclear what type of information should be provided in driving scenarios of varying complexity and how this affects user trust and intention to use AVs. This study conducted a simulated driving experiment (N = 32) using a 2 (Transparency Direction: Unidirectional Transparency vs. Bidirectional Transparency) × 2 (Transparency Level: Low vs. High) × 2 (Driving Scenario Complexity: Complex vs. Simple) within-subject design. User trust and intention to use AVs under different transparency conditions were evaluated. The results indicated that, in complex scenarios, bidirectional transparency in human-machine interaction increased user trust. Higher transparency levels enhanced both trust and intention to use AVs. Trust was found to fully mediate the effect of transparency on the intention to use. These findings offer valuable insights into optimizing the transparency design of AVs.