The aim of this study is to investigate if children's food preferences can be reliable measured by using pictures of foods presented on a computer screen in a conjoint layout. We investigate reproducibility (test–retest) and infer validity by comparison with traditional hedonic evaluations of tangible products and a choice test. In total, 300 children from schools in Copenhagen participated in the study. Two product cases were included: 32 buns and 8 juices. Pictures of the products were presented and evaluated on a computer screen using a conjoint layout. Right after the test, four buns and two juices (tangible products), chosen to span the preference spectrum, were hedonically evaluated for appearance and taste. Finally, an actual product choice was performed by having the children choose between two buns and two juices. Results showed that the computer evaluations with pictures of foods provided reproducible information about the children's visual food preferences, which were in concordance with both hedonic measures and products choices, and can thus be considered valid.