Here we show that the low temperature phase of magnetite is associated with an effective, although fractional, ordering of the charge. Evidence and a quantitative evaluation of the atomic charges are achieved by using resonant x-ray diffraction (RXD) experiments whose results are further analyzed with the help of ab initio calculations of the scattering factors involved. By confirming the results obtained from X-ray crystallography we have shown that RXD is able to probe quantitatively the electronic structure in very complex oxides, whose importance covers a wide domain of applications. Known in ancient times as lodestone and used to magnetize the mariner's compass [1], magnetite (Fe 3 O 4 ) is still today the archetype compound of a number of physical properties and applications. For instance, it is a promising candidate for the development of highly sensitive magneto-resistive devices in spin electronics. A sufficient comprehension of magnetite is mandatory in view of its applications and as a reference for similar effects in related materials. Particularly, the magnetic and magneto-electric properties of magnetite exhibit a conspicuous anomaly at T V = 121K that still defies understanding [2]. First suggested by Verwey [3] and expected in many oxides [4,5], the low temperature phase transition in magnetite has been associated to a charge disproportion on the metal atoms sites although a direct confirmation has never been evidenced in magnetite up to now. Theoretical predictions [6,7] have recently supported Verwey's scheme of the localization of the charge (but on a more complex pattern) and nuclear magnetic resonance [8] and Mössbauer [9] experiments are compatible with different oxidation states of the octahedral iron sites. However direct confirmation of charge disproportionation is still lacking.From the structural point of view the determination of the atomic positions issued from the metal-insulator transition challenges the scientific community ever since Verwey's seminal work [2,3]. Progress in the solution of the problem runs parallel to the development of new and sophisticated experimental techniques as well as to the implementation of refined computing codes and appropriate data analysis strategies. Despite all these technical advances, magnetite still remains a rather difficult case for conventional crystallography: the symmetry lowering (F d3m → Cc, and a c × a c × a c →≈ √ 2a c × √ 2a c × 2a c , with a c =8.394Å) generates 8 and 16 non equivalent iron sites at tetrahedral and octahedral positions, respectively, each one with its own atomic charge. However, the final structure is not yet perfectly known. The actually best refinement has been recently performed by Wright and collaborators [10] (in space group P mca, with lat-Complexity is greatly reduced in this structure model (Fig. 1), where there are 6 non equivalent iron atoms, two in tetrahedral sites (Fe t ) and four in octahedral sites (Fe 1 , Fe 2 , Fe 3 and Fe 4 ). Only the octahedral irons are supposed to undergo a charge ordering (CO)...