<p>Additive manufacturing (AM) promises more freedom in design than ever with maintaining consistent and reliable mechanical behavior across processes and materials being a significant challenge. Recent discoveries in machine learning (ML) provide effective data assistance to forecast mechanical properties e.g., tensile strength, porosity, and hardness by revealing non-linear process-structure–property interrelationships. This is a systematic review of the developments published since January 2021 until June 2025, according to the guidelines of PRISMA, in order to assess the status of ML-based property prediction in AM. Thirty peer-reviewed articles were reviewed in metals, polymers, and composites manufactured through such processes as laser powder bed fusion, directed energy deposition, fused deposition modelling, and vat photopolymerization. We benchmark widely used algorithms like artificial neural networks, support machine learning, random forests, and new hybrid or physics inspired models with regard to their data demands, validation schemes, predictive capability and constraints. The specific focus is made on the lack of datasets, the generalizability of models, their interpretability, and their connection with on-site monitoring. The review also suggests a conceptual pipeline of integrating ML into AM pipelines, which covers data acquisition, feature engineering, model training, and deployment. Shedding light on how ML can be used to enhance predictive reliability, quality assurance, and process optimization in additive manufacturing, this article can inform researchers and practitioners with systematic information on the recent theory and support the development of the research field by bridging research gaps.</p>