This study discloses the effect of hydrogen impurities on the local chemical bonding states and structure of amorphous alumina films by predicting measured Auger parameter shifts using a combination of atomistic and electrostatic modeling. Different amorphous alumina polymorphs with variable H-content and density, as grown by atomic layer deposition, were successfully modeled using a universal machine learning interatomic potential. The annealing of highly defective crystalline hydroxide structures with experimental H-contents at the corresponding atomic layer deposition temperatures led to excellent agreement between theory and experiment in the density and structure of the resulting amorphous alumina polymorphs. The measured Auger parameter shifts of Al cations in such polymorphs were accurately predicted with respect to the H content by assuming that all H atoms are present in the form of hydroxyl ligands in the randomly interconnected 4-fold, 5-fold, and 6-fold nearest-coordination spheres of Al. As revealed by a combination of atomistic and electrostatic modeling, the measured Auger shifts with an increase in the H content and an accompanying decrease in the oxide density depend on the complex correlations between local coordination, bond lengths, bond angles, and ligand type(s) around the core-ionized atoms. Moreover, cryogenic X-ray photoelectron spectroscopy is suggested to offer new insights into the local chemical and structural building blocks of crystalline and amorphous oxides by reducing thermal noise. These findings and fundamental knowledge contribute to advancing the design of e.g. hydrogen oxide barrier films, oxide membranes for H separation, H storage materials, and fuel cells for a hydrogen-based economy.