AbstractAccurate prediction of extreme weather and climate is crucial for their devastating impacts on our society and environment. Here, predictability of frequency of summer (June–August) extreme hot days (SEHDs) in China has been assessed in the SINTEX‐F2 seasonal forecast system during the period of 1983–2015. The hindcast had 12 ensemble members and was initialized on the first day of March, April and May, respectively. Results show that overall, the SINTEX‐F2 predicts more regions with good prediction skills at shorter lead time due to its better capture of the linear trends. Whereas, in Southwest China and eastern Tibetan Plateau, the prediction skills are consistently increased at shorter lead time even without the impacts of linear trends; the correlation coefficients between the region‐mean anomalies in the observed and predicted SEHD frequencies are 0.74, 0.68 and 0.61 at 1–3 month lead, respectively, and they remain as high as 0.56, 0.49 and 0.40 when the linear trends are removed. This is because the SINTEX‐F2 can reproduce the observed influences of Indian Ocean Basin Mode (IOBM) on the SEHD frequency. The warm IOBM can cause anomalous ascending and strong divergence in the upper troposphere over the tropical Indian Ocean and Indian subcontinent. The strong divergence causes direct convergence and descending over Southwestern China and eastern Tibetan Plateau. It also causes substantial descending over the western North Pacific that can introduce anomalous cyclonic circulation over northeastern Asia via the Pacific‐Japan or East‐Asia pattern. The latter serves as a source of wave energies, excites the Rossby waves meandering westward in the mid‐latitude, and enhances the anticyclonic circulation anomalies over Southwestern China and eastern Tibetan Plateau.