Machine learning holds immense potential to accelerate nanoparticle drug delivery design by predicting complex in vivo behaviours. However, the evidence base has not been systematically reviewed, limiting understanding of progress and gaps. This systematic review maps and critically appraises the application of machine learning models to predict drug release, biodistribution, toxicity, and targeting for nanoparticle delivery systems from 2017 to 2026. The review focuses on model inputs, algorithms, validation strategies, and translational relevance. A PRISMA-compliant search of three databases identified 30 eligible studies. Data on ML techniques, nanoparticle types, outcomes, and validation methods were extracted and assessed for quality. Random forest, support vector machines, and deep neural networks dominated, with increasing use of graph-based and artificial intelligence-guided design approaches. Most studies focused on release prediction, while biodistribution and targeting models were less common. While ML in nanomedicine is growing rapidly, significant methodological gaps remain. The review highlights critical needs for standardized data, rigorous validation, and model interpretability to enable clinical translation.