Integrating visual and auditory information during gesture-speech integration (GSI) is important for successful social communication, which is often impaired in schizophrenia. Several studies suggested the posterior superior temporal sulcus (pSTS) to be a relevant multisensory integration site. However, intact STS activation patterns were often reported in patients. Thus, here we used Dynamic Causal Modelling (DCM) to analyze whether information processing in schizophrenia spectrum disorders (SSD) is impaired during GSI on network level. We investigated GSI in three different samples. First, we replicated a recently published connectivity model for GSI in a healthy subject group (n = 19). Second, we investigated differences between patients with SSD and a matched healthy control group (n = 17 each). Participants were presented videos of an actor performing intrinsically meaningful gestures accompanied by spoken sentences in German or Russian, or just telling a German sentence without gestures. Across all groups, fMRI analyses revealed similar activation patterns, and DCM analyses resulted in the same winning model for GSI. This finding directly replicates previous results. However, patients revealed significantly reduced connectivity in the verbal pathway (from left middle temporal gyrus (MTG) to left STS). The clinical significance of this connection is supported by its correlations with the severity of concretism and a subscale of negative symptoms (SANS). Our model confirms the importance of the pSTS as integration site during audio-visual integration. Patients showed generally intact connectivity during GSI, but revealed impaired information transfer via the verbal pathway. This might be the basis of interpersonal communication problems in patients with SSD.