earcons) on driving trust and takeover efficiency from two perspectives of takeover efficiency and trust scales. Considering that high cognitive load could extend the driver's takeover reaction time, a cognitive load scale was also used to assist in measuring takeover efficiency. Firstly, compared to the condition without auditory alert, both semantic and non-semantic alert types significantly enhanced driver's trust in the autonomous driving system. Secondly, among the semantic alert types, auditory icons performed better in takeover efficiency than voice warnings (P=0.000<0.001). Even with compression applied to both voice warnings and auditory icons, compressed auditory icons still outperformed compressed voice warnings in takeover efficiency (P=0.000<0.001). Finally, non-semantic alert demonstrated better takeover efficiency than semantic alert, with Abstract earcons showing faster takeover response than semantic alert in this experiment. Introducing auditory alert during driving takeovers is an effective strategy to enhance driving trust and reduce cognitive load. Therefore, autonomous driving systems should optimize the type and presentation method of alert to improve driver's takeover response time, ensuring system efficiency and enhancing the user's driving experience.