Previous neuroscience studies have provided important insights into the neural processing of third-party social interaction recognition. Unfortunately, however, the methods they used are limited by a high susceptibility to noise. Electroencephalogram (EEG) frequency tagging is a promising technique to overcome this limitation, as it is known for its high signal-to-noise ratio. So far, EEG frequency tagging has mainly been used with simplistic stimuli (e.g. faces), but more complex stimuli are needed to study social interaction recognition. It therefore remains unknown whether this technique could be exploited to study third-party social interaction recognition. To address this question, we first created and validated a wide variety of stimuli that depict social scenes with and without social interaction, after which we used these stimuli in an EEG frequency tagging experiment. As hypothesized, we found enhanced neural responses to social scenes with social interaction compared to social scenes without social interaction. This effect appeared laterally at occipitoparietal electrodes and strongest over the right hemisphere. Hence, we find that EEG frequency tagging can measure the process of inferring social interaction from varying contextual information. EEG frequency tagging is particularly valuable for research into populations that require a high signal-to-noise ratio like infants, young children and clinical populations.