Solvent effects on 31P‐NMR parameters for triphenylphosphine oxide and triphenylphosphine in chloroform have been extensively investigated by testing different solvation models. The solvent is described implicitly, mixed implicitly/explicitly, and using full explicit models. Polarizable continuum model (PCM), molecular dynamic (MD) simulations, and hybrid quantum mechanics/molecular mechanics (QM/MM) calculations are used to disclose the effects of solute/solvent interactions and, more generally, the role of the embedding in NMR simulations. The results show the beneficial effect of carrying out QM/MM optimizations on top of geometries directly extracted from classical MD simulations, used to ensure representative conformational sampling. The nuclear shielding convergence has been tested against a different number of snapshots and with the inclusion of solvent shells into the QM region. An automated MD//QM/MM//GIAO protocol, implemented in the COBRAMM package, is here proposed and tested on trimethyl phosphite showing that our approach boosts the convergence of nuclear shielding satisfactorily. The present work aims to be a stepping‐stone to assess proper QM/MM computational strategies in simulating chemical shifts in non‐homogeneous systems like supramolecular and biological systems.