It is well known that annihilations in the homogeneous fluid of dark matter (DM) can leave imprints in the cosmic microwave background (CMB) anisotropy power spectrum. However, the relevance of DM annihilations in halos for cosmological observables is still subject to debate, with previous works reaching different conclusions on this point. Also, all previous studies used a single type of parameterization for the astrophysical reionization, and included no astrophysical source for the heating of the intergalactic medium. In this work, we revisit these problems. When standard approaches are adopted, we find that the ionization fraction does exhibit a very particular (and potentially constraining) pattern, but the currently measurable τreio is left almost unchanged: in agreement with most of the previous literature, for plausible halo models we find that the modification of the signal with respect to the one coming from annihilations in the smooth background is tiny, below cosmic variance within currently allowed parameter space. However, if different and probably more realistic treatments of the astrophysical sources of reionization and heating are adopted, a more pronounced effect of the DM annihilation in halos is possible. We thus conclude that within currently adopted baseline models the impact of the virialised DM structures cannot be uncovered by CMB power spectra measurements, but a larger impact is possible if peculiar models are invoked for the redshift evolution of the DM annihilation signal or different assumptions are made for the astrophysical contributions. A better understanding (both theoretical and observational) of the reionization and temperature history of the universe, notably via the 21 cm signal, seems the most promising way for using halo formation as a tool in DM searches, improving over the sensitivity of current cosmological probes.