<p>Soil quality is controlled by a vast number of biological, physical, and chemical factors, the proper evaluation of these factors is essential to the sustainability of agricultural production and the efficient management of resources. Conventional methods for soil assessment are still labour intensive, time consuming and highly dependent on expert interpretation. Contemporary advancements in internet of things (IoT) and machine learning (ML) have led to the creation of automated and intelligent systems of monitoring soil that improve the accuracy of forecasting while simultaneously reducing the amount of manual input. This work provides a systematic review of IoT empowered ML methods for the evaluation of soil quality parameters. The review followed the guidance of the preferred reporting items for Systematic Reviews and Meta-Analyses (PRISMA) using the major scientific databases (Scopus, Web of Science, IEEE Xplore, Science Direct and Google Scholar) to capture the studies published between 2010 and 2025. The novelty of the review is an integrated structure through which sensor technologies, machine learning performance, and evaluation metrics of IoT-supported soil and agricultural surveillance are studied. The selected literature is classified based on a taxonomy, soil parameters (eg, moisture, pH, and nutrients), deployment environments, and learning paradigms, supervised, ensemble, and deep learning techniques. Three research questions were addressed focusing on the strategies of sensor deployment, algorithmic efficacy, datasets, evaluation metrics, and key challenges documented in the literature. The results provide an understanding of both current achievements and open questions related to data quality, sensor reliability, model generalization and real-world deployment, and therefore have value both as insights and as potential avenues for future research for both scholars and practitioners.</p>