[en] Gaia will provide astrometric and photometric observations for about 500 000 quasars distributed over the whole sky. The latter would constitute an isotropic grid of fixed sources perfectly suited to determine the referential frame. However, they must first be properly identified among stars, whose population is about 2000 times larger. Using broad and medium band photometry, we first compare the efficiency of two analysis methods (Ï 2 fitting and Artificial Neural Networks, ANNs) to produce the QSO catalogue with the lowest amount of contamination by stars. We then investigate whether the Gaia photometry could also provide precise values of the QSO astrophysical parameters (APs, like the redshift, the continuum slope, the emission line strength and possibly the extinction). To reach that purpose, we also compare the performances of the Spectral Principal Component Analysis (SPCA) with those of the Ï 2 fitting and ANN analysis.