This study focuses on the interactive guidance design methods in metaverse education and training systems, analyzing the main factors affecting learners’ cognitive load levels and training learning outcomes in the metaverse education and training process. The study classifies six interactive guidance elements into two groups based on the amount of information conveyed from visual and auditory dimensions and sets them in two training modes: learning and practice. Using a 2 × 2 multifactorial controlled experiment, the study analyzes test scores, test completion times, and subjective cognitive load scale data of participants to examine the differences in training effectiveness and cognitive load levels among learners with different prior experiences in training modes composed of different interactive guidance elements. The experimental results indicate that the high-experience group outperformed the low-experience group in test performance across both training modes. For high-experience learners, there was no significant difference in test performance after training in both modes, but learners trained in the learning mode had slightly lower cognitive load levels than normal during the training process.