Ultra-compact X-ray binaries are identified by their extremely short orbital periods of less than 1h. This implies such tight orbits that only H-deficient compact objects would fit. Most likely, UCXBs consist of a Roche lobe filling white dwarf that is accreting onto a neutron star companion. This extreme environment offers a unique opportunity to peek at the interior of compact stars and the accretion of H-deficient material, as well as exploring questions of mass-transfer stability and the physics of compact object mergers. Depending on the evolutionary scenario leading to their creation, the donor could be a helium star. a He white dwarf or a C/O white dwarf. Determining the chemical composition of the disk (and therefore the donor star) in UCXBs is a crucial step in improving our understanding of the physics behind their origin and evolution. In our recently published paper we show that X-ray reflection spectra, in particular the iron Ka line, can be used as a diagnostic of the chemical composition of the accretion disk in UCXBs. Namely, we use Monte Carlo simulations to show that the most dramatic and easily observable consequence of a C/O-rich accretion disk is the more than tenfold attenuation of the Ka line of iron (consistently detected in the spectra of normal LMXBs with main sequence donors). On the other hand in the case of a He-rich donor the iron line remains at its usually observed strength. In my talk I will be presenting the details of our theoretical model and its predictions, as well its application to the spectra of well known UCXBs (Koliopanos et al. 2013 in prep.) and our conclusions regarding their chemical composition.