Abstract We present the direct detection of the splashback feature using the sample of massive galaxy clusters from the Local Cluster Substructure Survey (LoCuSS). This feature is clearly detected (above 5σ) in the stacked luminosity density profile obtained using the K-band magnitudes of spectroscopically confirmed cluster members. We obtained the best-fit model by means of Bayesian inference, which ranked models including the splashback feature as more descriptive of the data with respect to models that do not allow for this transition. In addition, we have assessed the impact of the cluster dynamical state on the occurrence of the splashback feature. We exploited the extensive multiwavelength LoCuSS data set to test a wide range of proxies for the cluster formation history, finding the most significant dependence of the splashback feature location and scale according to the presence or absence of X-ray emitting galaxy groups in the cluster infall regions. In particular, we report for the first time that clusters that do not show massive infalling groups present the splashback feature at a smaller clustercentric radius r sp/r 200,m = 1.158 ± 0.071 than clusters that are actively accreting groups r sp/r 200,m = 1.291 ± 0.062. The difference between these two subsamples is significant at 4.2σ, suggesting a correlation between the properties of the cluster potential and its accretion rate and merger history. Similarly, clusters that are classified as old and dynamically inactive present stronger signatures of the splashback feature, with respect to younger, more active clusters. We are directly observing how fundamental dynamical properties of clusters reverberate across vastly different physical scales.