Reflection and high-energy cut-off are with intrinsic absorption the main spectral properties of AGNs. Studying them is, however, not as simple as studying intrinsic absorption. This is because both reflection and high-energy cut-off produce a weaker spectral signature and with low signal to noise data they tend to be degenerate. The optimum solution would be to use high-quality broad band data which allow to disentangle all three spectral components. Such approach has been recently used for type-1 and type-2 AGNs by [59] and [60] respectively. Both groups find evidences for spectral cut-offs which are at energies < 300 keV and thus at odds with previous studies in the 1-500 keV range [61, 62] which located the cut-off at > 500 keV. In most cases a strong reflection component and an Iron line are found, although they appear to be inconsistent with the measured absorbing column densities [60]. This might indicate that the absorbing and the reflecting material are not the same or that the absorbing/reflecting material is not uniform (i.e. a clumpy torus model). [63] analyzed a sample of 105 Seyfert objects detected by BeppoSAX and thus with simultaneous broad-band coverage in the 2-100 keV band. He finds a systematic difference between Seyfert 1 and 2 galaxies with Seyfert 1s having a steeper spectrum, a larger reflection component and a lower energy spectral cut-off. This is in agreement with what found for a smaller sample of Swift/BAT sources [31]. While the different strengths of the reflection components might be explained by the different viewing angles of the reflecting material in these objects, the different spectral cut-offs might signal a break down of the AGN unified model. Indeed, in thermal Compton models (e.g. [61]) the cut-off energy is related to the thermal energy of the electrons populating the hot corona above the accretion disk and in this case it would imply that this energy is lower in Seyfert 1 with respect Seyfert 2. However, I remark that with low signal to noise spectra is very difficult to disentangle all the different spectral components. The best way would be to select a flux-limited sample of bright Seyfert galaxies for which a 0.1-500 keV data exist. While thanks to XMM-Newton, Swift and INTEGRAL it is possible to assemble high-quality data up to ~ 200 keV, above this energy there are essentially no data. In this framework the detection of Seyfert galaxies above 200 keV (e.g. [64]) might play an important role as it would allow to disentangle the spectral cut-off from the other spectral features.