This study aimed to investigate the association of ultra-processed food consumption at 4 and 7 years of age with appetitive traits at 7 years and BMI at 10 years of age. Participants were 1175 children of the population-based birth cohort Generation XXI, who provided food diaries and complete data on socio-demographic variables, anthropometric measures and the Portuguese Children’s Eating Behaviour Questionnaire (P-CEBQ). Foods were grouped according to NOVA classification into: ‘unprocessed, minimally or moderately processed, and culinary preparations’; ‘processed’ and ‘ultra-processed’. To assess tracking of groups’ consumption, Pearson’s correlation coefficient (r) and the intraclass correlation coefficient (ICC) were calculated. Generalised linear models were fitted to test main associations, mediators and interactions among the variables. Ultra-processed consumption exhibited a fair level of stability between ages 4 and 7 years (r 0·34; ICC = 0·32; 95 % CI 0·25, 0·39), corresponding, respectively, to 27·3 % (1881·9 (SD 908·8) kJ/d) and 29·3 % (2204·5 (SD 961·1) kJ/d) of total energy intake. After adjusting for maternal and child characteristics, higher ultra-processed consumption at 4 years was associated directly with ‘Food Responsiveness’ (β = 0·019; 95 % CI 0·007, 0·037) and indirectly through energy intake with avoidant traits: ‘Food Fussiness’ (β = –0·007; 95 % CI 0·002, 0·012) and ‘Satiety Responsiveness’ (β = –0·007; 95 % CI 0·003, 0·012). Ultra-processed consumption at 4 years old was associated with BMI at 10 years old, but appetitive behaviours were not powerful mediators of this association. The results suggest a path by which ultra-processed products may impact on later appetitive traits and higher BMI in children.