We have studied the sampling performance of conformational search programs using geometric and energetic criteria. Ideally, a conformational search algorithm should identify the largest possible number of low-energy structures (energy criterion) covering the widest possible range of molecular shapes (geometric criterion). Geometric analysis consisted in comparing the distribution of conformations within the generated ensembles by multidimensional scaling and by analysing the eigenvalue structure of the pairwise coordinate covariance matrices. The energetic comparison was carried out by assessing the energy distribution of conformers after minimizing them all using the same semi-empirical quantum mechanics optimization protocol. The present investigation focused on five conformational search programs: DGEOM, QXP, ROTATE, LMOD and OMEGA. We have applied these methodologies to a maximally diverse 604-compound subset of the LeadQuest library. The program LMOD performs best according to the energetic criterion, whereas a wider range of geometrically diverse conformations is sampled by the other programs, at the cost of higher median conformer energies. In terms of speed, OMEGA is fastest. We recommend the use of LMOD or OMEGA for high-quality conformational search applications.