Mediterranean marine protected areas require repeatable fish monitoring to evaluate conservation and fisheries outcomes, yet many management programmes begin with sparse, heterogeneous observations. This study evaluates a pilot underwater visual census undertaken on 19 August 2026 in Al Hoceima National Park, Morocco, places it within surveys conducted in 2002 and 2019, and uses the combined evidence to design an operational sentinel-monitoring framework. Two contrasting coastal units - one recorded as natural habitat and one as an artificial breakwater - were surveyed in 2026. Fish counts were consolidated by taxon and field size class, taxonomic names were quality-controlled, and assemblages were described using abundance, richness, Shannon diversity, Pielou evenness, Simpson dominance and Bray-Curtis dissimilarity. The 2002 MedMPA survey recorded 72 species across multiple sites and depth bands, whereas 22 visual surveys in 2019 recorded 3,577 individuals and 16 families; reported richness varies between 42 and 44 species within the source report. Those historical datasets identify recurrent sparid and labrid components and high proportions of small fish at some shallow locations, but their different coverage and effort preclude a direct numerical trend with 2026. Because the 2026 transect geometry, depth and spatial replication were not documented, analyses were intentionally restricted to raw-count descriptions and no habitat effect was tested. The pilot recorded 290 individuals representing nine taxonomic units. The natural unit contained 199 individuals and seven taxa, whereas the breakwater contained 91 individuals and five taxa. Four schooling or abundant groups - Boops boops, Sarpa salpa, Oblada melanurus and Mugil sp. - accounted for 90.0% of all records. Bray-Curtis dissimilarity was 0.745, indicating a marked descriptive contrast that cannot be attributed causally to habitat. The pilot supplies a traceable summer baseline, while the historical surveys guide station placement and ecological questions. Fixed georeferenced transects, replicated habitat units, standardized effort, environmental covariates, photographic validation and seasonally homologous sampling are required before interannual change can be estimated.
