As a new intrusion method in the security field, phishing poses an enormous threat to network security and personal privacy. Thus, improving the level of network security and preventing phishing are a matter of great concern to both the state and researchers. A 2 (automation trust tendency) *2 (system reliability level) *2 (feedback) between-subjects design was adopted to study the impact of individual characteristics and system features on phishing detection. Three hundred ninety-eight participants completed a phishing email task to identify whether 40 emails were legitimate or fraudulent. The results showed that systems with feedback and high reliability improve users’ performance in email identification. Users with a high tendency towards automation trust have a higher risk of phishing. However, feedback from the system helps calibrate a high automation trust tendency. These research results can promote an understanding of phishing prevention mechanisms and provide support for the design of email defence systems.