Results coming from the X-ray stacking, showing a significant X-ray weakness of all three samples of NLAGNs at z < 4.5, NLAGNs at z > 4.5, and high-z BLAGNs. Left: Expected X-ray luminosity versus observed X-ray luminosity (in the 2–10 keV band) of the three AGN samples we used for the stacking, as labeled. We also plot the sources in the CEERS MR sample with an X-ray detection color-coded by their obscuration level, as derived from the X-ray spectral analysis (Buchner et al. 2015). Diamonds refer to observed X-ray luminosities, while circles refer to the intrinsic X-ray luminosities derived from the X-ray analysis. Errors bars mark the error on the intrinsic X-ray luminosities of the sources reported in Buchner et al. (2015). The result from the X-ray stacking of these X-ray detected sources is reported as a gold diamond, while the gold circle considers the median of the intrinsic X-ray luminosities of the X-ray sources. The dashed lines correspond to the shift from the observed to intrinsic X-ray luminosities. In cyan, we report the z = 1.037 BLAGN CEERS-3129, presented in Appendix A. Right: Ratio between the mean bolometric luminosity of the three AGN samples (derived using the values computed in Sect. 4.7) and the observed X-ray luminosity upper limits versus the mean bolometric luminosities. The orange line shows the reference L_{2 − 10 keV} − L_bol relation derived by Duras et al. (2020) and used to derive the expected X-ray luminosity of the sources. The empty hexagons represent the lower limits derived from the X-ray stack of the JWST-selected BLAGNs (red) and NLAGNs (green for z > 3 and blue for z < 3) on the CDFS and CDFN (Maiolino et al. 2025).