Background: Declining global biodiversity is driving efforts to find effective approaches to aid ecosystem conservation and management. New tools to rapidly and accurately gather biodiversity data are necessary for informed management. Metabarcoding of environmental DNA (eDNA) allows the simultaneous identification of multiple species from DNA present in environmental samples without biological source material. This technique offers the possibility of monitoring substantial components of biodiversity in a non-invasive, economical and timely manner. An eDNA approach could help marine ecosystems, due to their reduced accessibility, cryptic species, and poorly known taxa. However, water movement between habitats through currents and tidal influences could transport DNA from one area to another, leading to false-positive species detection and inaccurate biodiversity data. Results: We examined the accuracy of the eDNA monitoring method in a marine setting by comparing the eDNA signal between two neighboring sites holding different community assemblages (rocky shore vs. sheltered mudflats, <1km distance). 64 species were identified from three amplicon targets. Community structure analysis found a clear difference in the retrieved eDNA signal between our sites and habitat preference of detected species showed little evidence of DNA transport between habitats. Conclusion: Our results prove the accuracy of the eDNA monitoring method for a coastal marine setting by showing a lack of evidence for DNA transport through water movement and the possibility of detecting habitat-specific species assemblages. Metabarcoding of eDNA could alleviate the problems of monitoring biodiversity in the marine environment, by accurately and quickly gathering the necessary data for ecosystem conservation and management.