A large spread in climate model predictions of warming in response to the radiative forcing by a doubling of CO₂, measured by the equilibrium climate sensitivity (ECS), can be largely ascribed to model uncertainties in representing low cloud shortwave (SW) radiative feedback. In multi-model simulations from the CMIP6 project, we show that the predicted ECS in a model is highly related to its depiction of seasonal variations of low cloud fractions (LCFs) over the extra-tropics in present-day simulations. Marked reduction of extratropical LCFs during the transition from winter to summer is found in high ECS models, but is largely absent in low ECS models. The rapid reduction of extratropical LCFs with increasing local surface temperature as manifested in the seasonal cycle in the high ECS models is also evident in the future climate, i.e., extratropical low clouds are significantly reduced under a warming climate, leading to a strong positive SW radiative feedback and thus a rapid warming trend. The strong reduction of extratropical LCFs with increasing surface temperature, however, is not supported by CALIPSO satellite observations, indicating that the high ECS predicted in these models is likely overestimated. The emergent relationship between the seasonal variations of extratropical LCFs and ECS derived in this study, with constraints based on satellite observations of seasonal LCF variability, yields an ECS value of 3.1K with a low-cloud contribution to standard error of ±0.1K— subject to the caveat that this is from a single emergent constraint that may omit other sources of uncertainty. Further analyses using a regime-based approach suggest that the distinct difference in the low-cloud feedback between the high and low ECS models can be ascribed to their prevailing cloud regimes over the extra-tropics.