<p>Understanding the greenhouse effect is conceptually challenging for elementary school students because it involves abstract mechanisms that are difficult to visualize. Developing students’ visual representations is therefore an important component of scientific literacy. However, many learners struggle to translate conceptual understanding into meaningful visual forms. This study examines changes in elementary students’ visual representations of the greenhouse effect following the implementation of the GenAI-Based Reading (GBR) learning model. A mixed-methods multiphase explanatory design was employed, involving 28 fourth-grade students at a private elementary school in West Java, Indonesia. The primary instrument was a Three-Tier Test measuring conceptual understanding, confidence levels, and visual representation skills. Students’ drawings were analyzed using a six-category rubric. The results indicate changes in students’ visual representations, reflected in a reduction of misconceptions and an increase in more meaningful representation categories. Statistical analysis using the Wilcoxon signed-rank test showed a statistically significant difference between pretest and posttest scores (<i>p</i> = 0.0256). However, most students remained in initial representation categories, indicating the need for further instructional support. Qualitative findings suggest that students’ representations remained dominated by macroscopic elements, indicating limited understanding of submicroscopic processes. The findings highlight the importance of pedagogical support in fostering students’ visual representations in AI-supported learning environments.</p>