This study aims to examine the use of Artificial Intelligence (AI) in science learning assessment by identifying the tools and techniques currently employed, the types of learning that can be effectively assessed, and the advantages and limitations associated with AI-based assessment practices in science education. A systematic literature review was conducted, analyzing 20 recent studies published between 2020 and 2025, focusing on AI applications in science learning assessment contexts. The findings reveal that AI effectively assesses various types of scientific learning, including conceptual understanding, practical skills, scientific argumentation, and problem-solving. Compared to traditional methods, AI offers significant advantages such as real-time feedback, personalized learning experiences, and reduced teacher workload. However, persistent challenges remain, including algorithmic biases, complexity of student responses, and ethical concerns. The thoughtful integration of AI into science learning assessment has the potential to transform education significantly. Nevertheless, careful attention must be paid to ensure its ethical and equitable use in educational settings.