The rapid advancement of artificial intelligence (AI) has opened new opportunities for developing interactive learning media in elementary education. This study aims to describe the level of learning activeness (keaktifan belajar) of sixth-grade elementary school students after participating in Natural and Social Sciences (IPAS) instruction using Gemini Canvas, a generative-AI feature capable of producing interactive HTML-based educational games, and to examine whether that level significantly exceeds the minimum criterion of the "active" category. A quantitative descriptive approach with a one-group posttest-only design was employed, involving 30 sixth-grade students (total sampling) who completed a self-report learning-activeness questionnaire immediately after the intervention. The questionnaire was adapted from a previously validated elementary-level learning-activity instrument built on Diedrich's eight indicators of learning activity. Data were analyzed descriptively (mean, standard deviation, category distribution) and inferentially using a Shapiro-Wilk normality test and a one-sample t-test against a criterion score of 65 (the lower bound of the "Active" category), complemented by a Cohen's d effect size. The results showed a mean posttest score of 73.57 (SD = 9.93), with 80.0% of students in the Active or Very Active category and none in the Less Active category. The one-sample t-test indicated that the mean score was significantly above the criterion (t = 4.725, df = 29, p < 0.001), with a large effect size (d = 0.863). These findings indicate that AI Gemini Canvas-based interactive game media is associated with a high level of student learning activeness. As a posttest-only design without pretest data, the conclusions are limited to the achieved level of activeness rather than the magnitude of change, and future studies are encouraged to include pretest measurement for direct gain analysis.