The dimerization of molecules in helium nanodroplets is known to preferentially yield structures of higher energy than the global energy minimum structure for a number of quite different monomers. Here, we explore dimerization in this environment using an atomistic model within statistically converged molecular dynamics (MD) trajectories, treating the solvent implicitly through the use of a thermostat, or more explicitly by embedding one monomer in a He100 cluster. The focus is on the two simplest carboxylic acids, formic and acetic, both of which have been studied experimentally. While the global minimum structure, which comprises two CO⋯HO hydrogen bonds, is predicted to be the most abundant dimer in the absence of the helium solvent, this is no longer the case once helium atoms are included. The simulations confirm the importance of kinetic trapping effects and also shed light on the occurrence of specific dynamical effects, leading to the occasional formation of high-energy structures away from minima, such as saddle configurations. Theoretically predicted infrared spectra, based on the MD statistics, are in good agreement with the experimental spectra.