<p>Understanding marine surface fish diversity is crucial for ecosystem management. However, the traditional sampling methods are often invasive, costly, or unsuitable for certain species or locations. Environmental DNA (eDNA) metabarcoding provides a non-invasive and relatively cheap alternative to explore patterns of diversity. It is important to recognize that, eDNA-based inference can vary across sampling methods, potentially impacting the validity of biodiversity assessments. To evaluate and compare the effectiveness of three eDNA sampling methods—ship-bottom intake (4.5&#xa0;m), Niskin bottles (5 or 10&#xa0;m), and bucket (0&#xa0;m)—for assessing fish diversity and fish community composition in the western North Pacific near Japan, we analyzed fish communities from 83 stations sampled during nine research cruises. Taxonomic analysis revealed that each method detected over 324 taxa, contributing to a total of 465 taxa. Hierarchical clustering generally identified similar species composition across methods at a station. The exception was when intake samples, collected at different times, diverged from bucket and Niskin samples at the same station. Hill’s number rarefaction and extrapolation curves across all clusters showed similar results among methods, with exceptions in a few clusters where bucket samples exhibited higher biodiversity indices than intake and Niskin samples. Non-metric multidimensional scaling indicated significant relationships between cluster composition and environmental factors like temperature, salinity, and chlorophyll-a. Some clusters were also controlled by integrated seasonal factors. Overall, fish community composition was convincingly similar among methods. This finding suggests that any of these eDNA sampling methods can be effective and may be prioritized based on logistical considerations.</p>