To face the current downward trajectory of freshwater biodiversity loss, the implementation of effective biodiversity monitoring programs is of utmost importance. Environmental DNA offers unprecedented opportunities for this aim but several challenges still need to be addressed before implementing efficient species monitoring using eDNA. One of them is optimizing the eDNA sampling scheme to maximize the eDNA detection probability. For instance, in flowing freshwaters, the transport of eDNA downstream from its source can impact the eDNA detection probability, and blur the link between eDNA detection and the local occurrence of the species. Here, we investigated the eDNA spatial range of Harttiella lucifer (Siluriformes, Loricariidae), a threatened neotropical siluriform fish inhabiting French Guianese mountain streams, and confined to waterfalls and fast‐flowing environments. Environmental samples were collected at 11 sites from the H. lucifer population to 2000 m downstream. A species‐specific dPCR approach was applied to quantify the amount of DNA present in each sample and evaluate the eDNA detection probability of H. lucifer according to the distance from its source. Results showed an accumulation of eDNA at 50 and 100 m downstream from H. lucifer population. The evaluation of detection probabilities revealed that 300 m downstream from H. lucifer population, the probability of detection drops to 50%. This study suggests that eDNA drift in neotropical small streams is limited to a few tens meters downstream. These findings demonstrate that in neotropical small streams, eDNA provides a picture of the local fish fauna rather than integrating information over large spatial scales.