Leptospirosis is the most prevalent bacterial zoonosis worldwide and, in this context, has been extensively investigated through the One Health framework. Diagnosis of human leptospirosis includes molecular and serological tools, with the serological Microscopic Agglutination Test (MAT) still being considered as the gold standard. Mammals acting as reservoirs of the pathogen include species or populations that are able to maintain chronic infection and shed the bacteria via their urine into the environment. Animals infected by Leptospira are often identified using the same diagnosis tool as in humans, i.e., serological MAT. However, this tool may lead to misinterpretations as it can signal previous infection but does not provide accurate information regarding the capacity of animals to maintain chronic infection and, hence, participate in the transmission cycle. We employ in this paper previously published data and present original results on introduced and endemic small mammals from Indian Ocean islands to show that MAT should not be used for the identification of Leptospira reservoirs. By contrast, serological data are informative on the level of exposure of animals living in a specific environment. We present a sequential methodology to investigate human leptospirosis in the One Health framework that associates molecular detection in humans and animals, together with MAT of human samples using Leptospira isolates obtained from reservoir animals occurring in the same environment.