Gravitational-wave standard sirens provide an independent probe of cosmic expansion since their luminosity distances are inferred directly from the gravitational-wave signal and, for bright sirens (BSs), the source redshifts are obtained through the identification of electromagnetic counterparts. In this work, we forecast the constraining power of future bright siren catalogues on the Hubble constant using simulated binary-neutron-star mergers detected by the Einstein Telescope with associated electromagnetic counterparts. We construct mock catalogues with different numbers of events and redshift distributions, with GW170817 as a reference BS, and analyse the resulting constraints within a flat $Λ$CDM cosmology. We find that, when bright sirens are used as a standalone probe, most of the information on $H_0$ is provided by low-redshift events, with the improvement in precision saturating for sources above $z \sim 1$; beyond that redshift, the constraining power of BSs is increasingly limited by the degeneracy with $Ω_m$, which prevents further gains in precision on $H_0$. In this case, approximately $90$ low-redshift BSs are required to reach $σ_{H_0} \sim 1\,{\rm km\,s^{-1}\,Mpc^{-1}}$, while about $45$ are sufficient for $σ_{H_0} \sim 2\,{\rm km\,s^{-1}\,Mpc^{-1}}$. When external BAO information is included to reduce the $H_0 - Ω_m$ degeneracy, intermediate-redshift sirens become more informative and the required number of events decreases substantially, to roughly $20$ and $15$ BSs for $σ_{H_0} \sim 1\,{\rm km\,s^{-1}\,Mpc^{-1}}$ and $σ_{H_0} \sim 2\,{\rm km\,s^{-1}\,Mpc^{-1}}$, respectively. These results highlight the importance of both electromagnetic counterpart identification and complementary background probes in making BSs a competitive, distance-ladder-independent test of the Hubble tension.