We investigated the reliability of the genetic algorithm which will be used to invert the photometric measurements of asteroids collected by the European Space Agency Gaia mission. To do that, we performed several sets of simulations for 10 000 asteroids having different spin axis orientations, rotational periods and shapes. The observational epochs used for each simulation were extracted from the Gaia mission simulator developed at the Observatoire de la Côte d'Azur, while the brightness was generated using a Z-buffer standard graphic method. We also explored the influence on the inversion results of contaminating the data set with Gaussian noise with different σ values. The research enabled us to determine a correlation between the reliability of the inversion method and the asteroid's pole latitude. In particular, the results are biased for asteroids having quasi-spherical shapes and low pole latitudes. This effect is caused by the low light-curve amplitude observed under such circumstances, as the periodic signal can be lost in the photometric random noise when both values are comparable, causing the inversion to fail. Such bias might be taken into account when analysing the inversion results, not to mislead it with physical effects such as non-gravitational forces. Finally, we studied what impact on the inversion results has combining a full light curve and Gaia photometry collected simultaneously. Using this procedure we have shown that it is possible to reduce the number of wrong solutions for asteroids having less than 50 data points. The latter will be of special importance for planning ground-based observations of asteroids aiming to enhance the scientific impact of Gaia on Solar system science.