In an online process-tracing study, we measured the actual usage of available back-of-pack nutrition information during substitutive food choices made by 240 participants who had the intention to eat healthy. Using mouse-tracking software in a computerized task in which participants had to make dichotomous food choices (e.g., coconut oil or olive oil for baking), we measured the frequency and time of nutritional information considered. Combined with demographic and psychosocial data, including information on the level of intention, action planning, self-efficacy, and nutrition literacy, we were able to model the determinants of inadvertent unhealthy substitutive food choices in a sequential multiple regression. Moreover, we found three groups of participants that clustered on energy, salt, sugar, and saturated fat information used.