Bader's atoms-in-molecules theory holds that a bond path between an atom pair indicates a bonded interaction, including when the interaction is intermolecular. There is now considerable evidence from a variety of sources that this proposition is unreliable, not least when the bond path is observed in an experimental electron density distribution. In any case, to conclude that a weak interaction is bonding does not show that the interaction will be useful in crystal engineering. For this it must also be competitive, i.e. likely to win out over other interactions that might form when a crystal nucleates and grows. Database techniques have been developed for identifying interactions that are generally competitive in organic structures, or competitive in specific types of structures. They have also been applied to protein–ligand complexes. Representative results are discussed. As would be expected, the most competitive interactions are usually the strongest but this is not invariably the case.