The evaluation of solvation energies is a great challenge. We focus here on an organic molecule chemisorbed at a metal-liquid interface, as a prototypical system, essential in tribology, electrochemistry and heterogeneous catalysis. We compare an established implicit solvation scheme with a strategy based on molecular mechanics (MM) free energy perturbation (FEP) seeded by QM computations. First, we benchmark the approaches against experimental hydration energies of standard (organic) molecules and find acceptable errors in the order of 0.06 eV (1.3 kcal mol-1). Then, the impact of various parameters on the solvation energy of an adsorbate have been assessed on a typical system of interest, levulinic acid adsorbed at a Ru(0001)/water interface. We identify the need for dipole corrections or symmetric slabs when including solvation effects on metallic surfaces. The MM-FEP scheme is revealed to be as reliable as the implicit solvent for water. In the case of levulinic acid, both PCM and MM-FEP agree that the bulk solvation effect is not sufficient to change the adsorption mode from bidentate to mono-dentate, despite the fact that the COOH group is desolvated in the bidentate case. MM-FEP has the great advantage of being more easily generalized to other solvents and to be further improved which will be particularly useful to study solvent and (counter-)ion effects on interfacial reactions.