Common problems in state-of-the-art climate models are a cold sea surface temperature (SST) bias in the equatorial Pacific and the underestimation of the two most important atmospheric feedbacks operating in the El Niño/Southern Oscillation (ENSO): the positive, i.e. amplifying wind-SST feedback and the negative, i.e. damping heat flux-SST feedback. To a large extent, the underestimation of those feedbacks can be explained by the cold equatorial SST bias, which shifts the rising branch of the Pacific Walker Circulation (PWC) too far to the west by up to 30°, resulting in an erroneous convective response during ENSO events. Based on simulations from the Kiel Climate Model (KCM) and the 5th phase of Coupled Model Intercomparison Project (CMIP5), we investigate how well ENSO dynamics are simulated in case of underestimated ENSO atmospheric feedbacks (EAF), with a special focus on ocean–atmosphere coupling over the equatorial Pacific. While models featuring realistic atmospheric feedbacks simulate ENSO dynamics close to observations, models with underestimated EAF exhibit fundamental biases in ENSO dynamics. In models with too weak feedbacks, ENSO is not predominantly wind-driven as observed; instead ENSO is driven significantly by a positive shortwave radiation feedback. Thus, although these models simulate ENSO, which in terms of simple indices is consistent with observations, it originates from very different dynamics. A too weak oceanic forcing on the SST via the positive thermocline, the Ekman and the zonal advection feedback is compensated by weaker atmospheric heat flux damping. The latter is mainly caused by a biased shortwave-SST feedback that erroneously is positive in most climate models. In the most biased models, the shortwave-SST feedback contributes to the SST anomaly growth to a similar degree as the ocean circulation. Our results suggest that a broad continuum of ENSO dynamics can exist in climate models and explain why climate models with less than a half of the observed EAF strength can still depict realistic ENSO amplitude.