In our study we analyse the online processing of visual-verbal input during simultaneous interpreting with text. To that end, we compared 15 professional interpreters’ eye movements during simultaneous interpreting with text (SIMTXT) to a baseline collected during reading while listening (RWL). We found that interpreters have a preference for a visual lead during RWL, following the pattern well-documented in silent and oral reading studies. During SIMTXT, in contrast, interpreters show a clear preference for a visual lag. We tentatively conclude that during SIMTXT the visual input might be used first and foremost to support the production of the output rather than the comprehension of the input. Importantly, we submit that the availability of the written text of the orally presented discourse might negatively affect predictive processing.