This study aimed to investigate the impact of AI-based formative assessment on improving the academic achievement of eighth-grade female students in science and their perceptions of the classroom assessment environment. The study employed a quasi-experimental design with pre-test and post-test for a control group and an experimental group. The sample consisted of 70 eighth-grade female students in Muscat Governorate. An AI-based formative assessment was used for the experimental group (n=34) for two units of the second semester, while the regular formative assessment was used for the control group (n=36). To achieve the study's objectives, two tools were used: first, a 15-item achievement test, whose psychometric properties were verified through content validity, item analysis, and reliability using Cronbach's alpha; second, a perceptions questionnaire about the classroom assessment environment, adapted from Alkharusi (2011), which included two main domains: learning-oriented assessment environment and performance-oriented assessment environment, each containing 9 items. The psychometric properties of the questionnaire were verified through content validity and reliability using Cronbach's alpha. To test the study hypotheses, an independent samples t-test was used for the achievement test, and multivariate analysis of variance was used for the perceptions questionnaire. The results showed statistically significant differences in both academic achievement and students' perceptions of the classroom assessment environment . For the benefit of the experimental group. This indicates the positive impact of AI-based formative assessment in raising the level of students' achievement in science, as well as promoting an effective and stimulating classroom assessment environment for students towards understanding and learning.