We present a group-galaxy cross-correlation analysis using a group catalog produced from the 16,500 spectra from the optical zCOSMOS galaxy survey. Our aim is to perform a consistency test in the redshift range 0.2 < z < 0.8 between the clustering strength of the groups and mass estimates that are based on the richness of the groups. We measure the linear bias of the groups by means of a group-galaxy cross-correlation analysis and convert it into mass using the bias-mass relation for a given cosmology, checking the systematic errors using realistic group and galaxy mock catalogs. The measured bias for the zCOSMOS groups increases with group richness as expected by the theory of cosmic structure formation and yields masses that are reasonably consistent with the masses estimated from the richness directly, considering the scatter that is obtained from the 24 mock catalogs. An exception are the richest groups at high redshift (estimated to be more massive than 10^13.5 M_sun), for which the measured bias is significantly larger than for any of the 24 mock catalogs (corresponding to a 3-sigma effect), which is attributed to the extremely large structure that is present in the COSMOS field at z ~ 0.7. Our results are in general agreement with previous studies that reported unusually strong clustering in the COSMOS field.