ABSTRACT We study the galactic bulges in the Auriga simulations, a suite of 30 cosmological magneto-hydrodynamical zoom-in simulations of late-type galaxies in Milky Way sized dark matter haloes performed with the moving-mesh code arepo. We aim to characterize bulge formation mechanisms in this large suite of galaxies simulated at high resolution in a fully cosmological context. The bulges of the Auriga galaxies show a large variety in their shapes, sizes, and formation histories. According to observational classification criteria, such as Sérsic index and degree of ordered rotation, the majority of the Auriga bulges can be classified as pseudo-bulges, while some of them can be seen as composite bulges with a classical component; however, none can be classified as a classical bulge. Auriga bulges show mostly an in situ origin, $21{{\ \rm per\ cent}}$ of them with a negligible accreted fraction (facc < 0.01). In general, their in situ component was centrally formed, with ${∼}75{{\ \rm per\ cent}}$ of the bulges forming most of their stars inside the bulge region at z = 0. Part of their in situ mass growth is rapid and is associated with the effects of mergers, while another part is more secular in origin. In $90{{\ \rm per\ cent}}$ of the Auriga bulges, the accreted bulge component originates from less than four satellites. We investigate the relation between the accreted stellar haloes and the bulges of the Auriga simulations. The total bulge mass shows no correlation with the accreted stellar halo mass, as in observations. However, the accreted mass of bulges tends to correlate with their respective accreted stellar halo mass.