Coupling a hydrophobic drug onto monoclonal antibodies via lysine residues is a common route to prepare antibody–drug conjugates (ADC), a promising class of biotherapeutics. But a few chemical modifications on protein surface often increase aggregation propensity, without a clear understanding of the aggregation mechanisms at stake (loss of colloidal stability, self-assemblies, denaturation, etc.), and the statistical nature of conjugation introduces polydispersity in the ADC population, which raises questions on whether the whole ADC population becomes unstable. To characterize the average interactions between ADC, we monitored small-angle X-ray scattering in solutions of monoclonal IgG1 human antibody drug conjugate, with average degree of conjugation of 0, 2, or 3 drug molecules per protein. To characterize stability, we studied the kinetics of aggregation at room temperature. The intrinsic Fuchs stability ratio of the ADC was estimated from the variation over time of scattered light intensity and hydrodynamic radius, in buffers of varying pH, and at diverse sucrose (0% or 10%) and NaCl (0 or 100 mM) concentrations. We show that stable ADC stock solutions became unstable upon pH shift, well below the pH of maximum average attraction between IgGs. Data indicate that aggregation can be ascribed to a fraction of ADC population usually representing less than 30 mol % of the sample. In contrast to the case of (monodisperse) monoclonal antibodies, our results suggest that a poor correlation between stability and average interaction parameters should be expected as a corollary of dispersity of ADC conjugation. In practice, the most unstable fraction of the ADC population can be removed by filtration, which affects remarkably the apparent stability of the samples. Finally, the lack of correlation between the kinetic stability and variations of the average inter-ADC interactions is tentatively attributed to the uneven nature of charge distributions and the presence of patches on the drug-modified antibodies.