Seagrass ecosystems, especially in small islands, play an important role in supporting coastal ecological functions for surrounding areas. Unfortunately, their condition and spatial distribution remain poorly documented in several protected areas within Jakarta Bay. This study aims to assess the ecological condition and spatial distribution of seagrass meadows in Rambut Island Wildlife Sanctuary (SMPR), Indonesia, through an integrated approach combining field survey and satellite-based mapping. Field data were collected using a transect-quadrat method at 8 stations to determine species composition, percent cover, and meadow condition based on Indonesian national standard. Seagrass mapping was conducted using Sentinel-2A imagery processed in Google Earth Engine (GEE) and classified using the Random Forest (RF) algorithm. Five seagrass species were identified: Cymodocea rotundata, Cymodocea serrulata, Halophila ovalis, Halophila spinulosa, and Thalassia hemprichii. Seagrass percent cover ranged 7.38–39.71%, with all stations classified as poor condition based on national standards. The total mapped seagrass area in 2025 reached 4.95 ha. Accuracy assessment shows an overall accuracy of 90.38%, with producer’s and user’s accuracies for the seagrass class of 88.57% and 96.88%, respectively. The results provide an updated baseline for seagrass condition and distribution in SMPR and demonstrate the applicability of integrating field-based ecological assessment with machine learning-based remote sensing for shallow coastal ecosystem monitoring.