Background: Problem-solving ability is an important skill for students to develop in the 21st century. Many previous studies have focused on the impact of diverse digital educational games on problem-solving ability. However, there is inconsistency in the findings, with some studies reporting positive effects of digital game-based learning (DGBL) on problem-solving ability, while others either found no significant improvement or reported negative effects. Objectives: The objective of this study was to conduct a three-level meta-analysis to investigate the influence of DGBL on problem-solving ability. Additionally, the study aimed to examine various moderating factors to explore the potential sources of inconsistency among the primary studies. Methods: Through a systematic literature search, 21 primary studies were identified, encompassing a total of 60 effect sizes. The analysis involved assessing the effectiveness of DGBL in enhancing students' problem-solving ability and examining heterogeneity among the studies. Moderator analysis was also conducted to explore the impact of different factors on the effectiveness of DGBL. Results and Conclusions: The results indicated that DGBL could effectively enhance students' problem-solving ability (g = 0.654), with significant heterogeneity observed among studies (Q = 858.358, p < 0.001). Furthermore, our findings suggested that studies focusing on the perception of problem-solving ability (g = 1.498) yielded significantly larger effect size compared to studies focused on problem-solving skill (g = 0.440). Moreover, combining relevant teaching strategies with DGBL (g = 1.246) resulted in a significantly better effect compared to using DGBL alone (g = 0.399). The analysis of moderating factors revealed that the effect size was significantly influenced by teaching strategy and problem-solving construct. Other examined moderating factors did not show significant effects on the inconsistent findings.