Red gold alloys owe their warm and attractive colour to the combination of gold and copper atoms. Thanks to their good mechanical properties and resistance to corrosion, red gold alloys are used in luxury industries as well as in dentistry. Around the equiatomic composition AuCu, the alloys undergo an ordering phase transformation A1->L10 at the origin of a significant hardness improvement. However, this phase transformation is also the cause of strong macroscopic distortions, which are reported in the literature since 1956. Although this transformation has been widely studied during the 20th century, these macroscopic distortions have not been understood so far and are still the cause of substantial losses in precision manufacturing. The aim of this thesis is to understand this phenomenon in order to prevent its occurrence. This is achieved by adopting an approach that is not usual for order/disorder transformations. The controversial displacive aspect of the transformation is indeed explored, besides its diffusional character. First, an industrial campaign is carried out to explore the conditions of appearance of the phenomenon. We used the hole drilling method to measure the influence of the residual stress on the macroscopic distortions. Besides, simple mechanical testing was performed in order to more precisely reproduce these deformations at the lab scale. The mechanical bending of thin lamellae revealed the occurrence of a new singular phenomenon during the phase transformation: a Thermally Activated Distortion with Amplification that we call TADA effect. This interesting effect was found to depend on the initial stress in the lamella before the phase transformation. It occurs even for very low stress state and causes a strong amplification of the bending, which continues beyond the imposed shape in a stress-free state. In parallel, an extensive microstructural analysis is performed in order to understand the link between the macroscopic behaviour and the phase transformation. In particular, the use of Electron Backscatter Diffraction (EBSD) allowed us to analyse the crystallographic variants of the L10 phase. A quantitative analysis of the variant selection was developed based on the maximal work criterion, already used in literature for deformation twinning and martensitic transformations. The large-scale quantification of the variant selection enabled to clearly establish the link between the microscopic lattice strain and the macroscopic shape of the sample. We conclude that the bending of the samples is initially caused by the stress assisted variant selection and continues in a stress free state due to the persistence of the initial variant selection. This persistence was also observed under opposite loading in uniaxial and bending tests. Most important, in certain conditions, this effect was found to generate a force. This highlights again the importance of the initial stress assisted variant selection as it determines the final texture and shape of the piece. This was well illustrated on additively manufactured red gold piece with strong residual stresses. Finally, the distortion effect could be observed and quantified in several industrial red gold alloys with different additional elements, which indicates that it relies on the general properties of the A1->L10 transformation. This thesis provides new insights on the mechanisms of diffusion-limited displacive transformations that are still subject to controversy.