This study presents a structured assessment of artificial intelligence (AI) methods for high-performance recycled aggregate concrete (HP-RAC) property prediction. The study addresses previous issues regarding methodological rigor by introducing a structured review process that includes information on data synthesis from selected peer-reviewed articles, database selection, and inclusion criteria. Adaptive neuro-fuzzy inference systems (ANFIS), support vector machines (SVM), decision trees, artificial neural networks (ANN), and evolutionary algorithms (EA) are among the AI models that are categorized and contrasted in the review according to their interpretability, prediction accuracy, and dataset needs. In order to improve model resilience and generalization, a focus is made on multi-output modelling, hybrid frameworks, and the incorporation of optimization techniques like genetic algorithms (GA) and particle swarm optimization (PSO). The report also emphasizes how new platforms like compressive strength prediction platform (CSPP), materials simulation toolkit for machine learning (MAST-ML), modelling of materials with deep learning (MODNet), help make predictions that are repeatable, scalable, and interpretable. To help with model selection in the workability, mechanical, and durability domains, a single benchmarking matrix is suggested. The review highlights important research needs, such as the requirement for better interpretability, consistent datasets, and integration with sustainability measures. By combining AI applications into a coherent benchmarking framework, this work makes a unique contribution and offers researchers and engineers useful information for improving HP-RAC performance and design.
