Species distributions models (SDM) that rely on estimated relationships between present environmental conditions and species presence-absence are widely used to forecast changes of species distributions caused by global warming but far less to reconstruct historical assemblages. By compiling historical fish data from the turn to the middle of the twentieth century in a similar way for several European catchments (Rhône, Danube), and using already published SDMs based on current observations, we: (1) tested the predictive accuracy of such models for past climatic conditions, (2) compared observed and expected cumulated historical species occurrences at sub-catchment level, and (3) compared the annual variability in the predictions within one sub-catchment (Salzach) under a future climate scenario to the long-term variability of occurrences reconstructed during an extended historical period (1800–2000). We finally discuss the potential of these SDMs to define a “reference condition”, the possibility of a shift in baseline condition in relation with anthropogenic pressures, and past and future climate variability. The results of this study clearly highlight the potential of SDM to reconstruct the past composition of European fish assemblages and to analyze the historical ecological status of European rivers. Assessing the uncertainty associated with species distribution projections is of primary importance before evaluating and comparing the past and future distribution of species within a given catchment.