Digital games are widely used in education to motivate students for science. Additionally, augmented reality (AR) is increasingly used in education. However, recent research indicates that these technologies might not be equally beneficial for students with different background characteristics. Moreover, students with different backgrounds may differ in their self-efficacy and interest when playing games and this could lead to differences in performance. Given the increased use of games and immersive technologies in education, it is important to gain a better understanding of the effectiveness of games for different student groups. This study focused on the role of students' socio-economic status (SES) and examined whether SES was associated with in-game performance and whether interest and self-efficacy mediated potential associations between SES and in-game performance. Since log data are increasingly used to predict learning outcomes and can provide valuable insights into individual behaviour, in-game performance was assessed with the use of log data. In total, 276 early secondary school students participated in this study. The results indicate that SES has no direct or indirect effect through self-efficacy and interest on in-game performance. However, a lower self-efficacy increased the likelihood to drop out of the game. These findings suggest that students from different socio-economic backgrounds are equally interested and self-efficacious while playing the game and that their performance is not affected by their background. The affordances of AR as an immersive learning environment might be motivating enough to help mitigate possible SES differences in students.