In this paper, an assessment is made regarding the effects of J–R curve fitting and extrapolation methods in two J‐integral criteria – namely crack initiation, Ji, and tearing instability, J50– which were obtained through the linear normalization technique. Power‐law, logarithmic and linear fits were concurrently applied to J–Δa data derived from sub‐sized compact tensile specimens machined from a nuclear grade steel and tested at 300 °C. Research results show that the logarithmic J–R fit is the most conservative approach within a broad range of elastic–plastic fracture resistance, compared to the conventional power‐law fit. On the other hand, the linear fitting method provided the most non‐conservative J‐predictions. The values of Ji and J50 have been successfully correlated with the net energy absorbed during Charpy impact testing of the materials.