Detection and quantification of pathogenic free-living amoebae (FLA) in water samples is critical for assessing water quality and for disease management issues. The most probable number (MPN) is commonly used to account for FLA in water. Nevertheless, this requires a high number of water replicates and working volumes, and a consequent number of non-nutrient agar (NNA)-plates seeded with Escherichia coli. Herein, we aimed at optimizing this difficult method, taking also into account key factors such as (i) the counting method, (ii) the delay between sample collection and sample processing, and (iii) the temperature during water sample transportation. To simplify the MPN method, we filtrated 1 × 1000 and 1 × 100 mL water samples, and cellulose acetate filters were cut in 10 parts and inverted on NNA-plates overlaid with E. coli. The comparison between the classical and our optimized MPN method showed that the final counts were similar, therefore validating the use of the optimized method. Our results also showed that for thermophilic FLA (such as Naegleria fowleri), water samples can be kept at around +30°C and processed within 24 h. This improved MPN method is now routinely used in our laboratory to control Naegleria sp. in the water samples in Guadeloupe.