Context. The line-of-sight (l.o.s.) velocity dispersion (σlos) of the stellar component of ultra-faint-dwarf galaxies (UFDs) typically ranges from 3 to 6 km s−1. Applying standard mass estimators, this hints at extreme dynamical mass-to-light ratios (M/L) of approximately ∼100─5000 M⊙/L⊙ within the half-light radius (r1/2), making UFDs the most dark-matter (DM) dominated galaxies known and critical tests for cosmological models. However, in this regime, it is a concern whether the l.o.s. velocity component of the orbital motion of undetected binary stars (binaries) is significantly inflating the observed σlos and, consequently, UFD dynamical mass estimates. Aims. Our goal is to correct the current estimates of σlos and dynamical masses of UFDs to account for the presence of undetected binaries with single-epoch data. Additionally, we generalized our methodology to work with multi-epoch data through which a fraction of stars forming part of binary systems can be detected via velocity variations. Methods. We used the latest binary population models in the solar neighborhood to compute the expected velocity distribution of binary stars. Then, we convolved this distribution with a Gaussian to model the l.o.s. velocity distribution of UFDs in a mixture model, in which the binary fraction is a free parameter. We applied this methodology to observed UFDs whose dynamical masses are potentially inflated by binaries. In order to generalize to the multi-epoch data case, we computed the velocity distribution of undetected binaries by applying the same cuts to the models that one would apply to the observed data to remove binaries. As the datasets currently available in the literature are not suitable for this, we only tested this method using a mock dataset. Results. We find that estimated dynamical masses of UFDs decrease by a factor of 1.5 to 3 once undetected binaries are accounted for. These corrections significantly affect considerations about DM models based on these systems. Additionally, the lower limits of the masses decrease significantly, even challenging the classification of Leo IV, Unions I, and Sagittarius II as galaxies. We demonstrate that a dedicated multi-epoch observational campaign spanning one year could substantially mitigate the impact of binaries, in particular if the presence of remaining undetected binaries is accounted for. Finally, we assessed the expected level of binary-star contamination in DM halo density profile inferences from dynamical models of classical dwarf spheroidal galaxies, and we find that it is negligible for Sculptor-like galaxies.