Abstract. The majority of photonuclear reaction cross sections important for many fields of science and technology and various data files (EXFOR, RIPL, ENDF, etc.) supported by the IAEA were obtained in experiments using quasimonoenergetic annihilation photons. There are well-known systematic discrepancies between the partial photoneutron reactions (γ , 1n), (γ , 2n), (γ , 3n). For analysis of the data reliability the objective physical criteria were proposed. It was found out that the experimental data for many nuclei are not reliable because of large systematic uncertainties of the neutron multiplicity sorting method used. The experimentally-theoretical method was proposed for evaluating the reaction cross sections data satisfying the reliability criteria. The partial and total reaction cross sections were evaluated for many nuclei. In many cases evaluated data differ noticeably from both the experimental data and the data evaluated before for the IAEA Photonuclear Data Library. Therefore it became evident that the IAEA Library needs to be revised and updated.