5G radio positioning exploits information in both angle and delay, by virtue of increased bandwidth and large antenna arrays. When large arrays are embedded in surfaces, they can passively steer electromagnetic waves in preferred directions of space. Reconfigurable intelligent surfaces (RIS), which are seen as a transformative "beyond 5G" technology, can thus control the physical propagation environment. Whereas such RIS have been mainly intended for communication purposes so far, we herein state and analyze a RIS-aided downlink positioning problem from the Fisher Information perspective. Then, based on this analysis, we propose a two-step optimization scheme that selects the best RIS combination to be activated and controls the phases of their constituting elements so as to improve positioning performance. Preliminary simulation results show coverage and accuracy gains in comparison with natural scattering, while pointing out limitations in terms of low signal to noise ratio (SNR) and inter-path interference.