Most tasks for measuring automatic approach-avoidance tendencies do not resemble naturalistic approach-avoidance behaviors. Therefore, we developed a paradigm for the assessment of approach-avoidance tendencies towards palatable food, which is based on arm and hand movements on a touchscreen, thereby mimicking real-life grasping or warding movements. In Study 1 (n = 85), an approach bias towards chocolate-containing foods was found when participants reached towards the stimuli, but not when these stimuli had to be moved on the touchscreen. This approach bias towards food observed in grab movements was replicated in Study 2 (n = 60) and Study 3 (n = 94). Adding task features to disambiguate distance change through either corresponding image zooming (Study 2) or emphasized self-reference (Study 3) did not moderate this effect. Associations between approach bias scores and trait and state chocolate craving were inconsistent across studies. Future studies need to examine whether touchscreen-based approach-avoidance tasks reveal biases towards other stimuli in the appetitive or aversive valence domain and relate to relevant interindividual difference variables.