In the present era of rapid global change, innovative monitoring methods can greatly enhance our ability to detect ecological disturbances and prioritise conservation areas in a timely and cost-effective manner. While Passive Acoustic Monitoring (PAM) has recently emerged as a promising tool for monitoring ecological states in marine environments, the specifics of how to apply this method remains poorly defined. In this study we examined how different combinations of sampling settings (frequency bandwidth, time of sampling (day/night), and sample duration) influenced the ability of two acoustic indices, the Sound Pressure Level (SPL) and the Acoustic Complexity Index (ACI), to discriminate different ecological states (ecostates) of coral reefs. We applied an iterative approach to select the most efficient and consistent combinations of sampling settings to use for these two acoustic indices, depending on the stability of their discriminating power across different time scales (successive days, moon phases, and seasons), and the minimum sampling effort required for reliable ecostate assessment. The ability of SPL and ACI to discriminate ecostate-specific soundscapes was more stable and required less sampling effort at nighttime. For indices calculated in the higher frequency band (>2¿kHz), very short recording times (=20¿min divided into 5¿s samples) were sufficient to discriminate ecostates, whereas longer recording times (=200¿min divided into 5¿min samples) were necessary when using indices calculated in the lower frequency bands (<1¿kHz). An optimised sampling scheme, i.e. the group of the five best combinations of settings to discern among coral reef ecostates, was determined at Reunion Island, Indian Ocean, then tested at New Caledonia, Pacific Ocean. Here, the classifications obtained through visual surveys and with the optimised acoustic sampling scheme were congruent. The concordance of our results with visual fish counts confirms the potential of ecoacousti