The Landsat programme recently launched Landsat 9 (L9). This sensor has not been widely evaluated for forest classification accuracies within New Zealand. This study evaluated L9 with two commonly used sensors, Landsat 8 (L8) and Sentinel-2 (S2). The creation of classifiers, imagery, and analysis was carried out using the cloud-based processing capabilities of Google Earth Engine (GEE), a cloud-based GIS software. The need for national small-scale forest classification system is of great importance. Improving the understanding of new sensors and their respective capabilities will improve their uptake within new systems and applications that may fulfil this need. The study was carried out in the southern part of the East Coast Region, below Gisborne City. Summer imagery was collected and processed for all three sensors. Imagery was integrated into a random forest classifier system, which was tested with ground truth data to attain different accuracy metrics for each sensor. S2 provided greater accuracy for forest classification with 95.0% accuracy when compared to L9 (93.7%) and L8 (93.7%). The Kappa co-efficient supported this with S2 (0.899), L8 (0.875), and L9 (0.875). McNemar’s test showed that S2 was performing significantly differently from L8 and L9 on the test dataset. From these three measures of accuracy and performance, it was concluded that L9 is not more accurate than S2 or L8 for forest classification in the East Coast of New Zealand. The key issue for L9 in this study was the classification of forest sizes less than 30 ha. Compared with S2, L9 struggled to accurately detect these areas and should not be implemented into systems that quantify resources with polygon sizes below this threshold (30 ha) if S2 is available. This was due to L9’s lower spatial resolution compared with S2, which likely caused more mixed value pixels, therefore increasing error.