The aim of this study was to determine whether food variety and perceived food preferences differ in infants following baby-led instead of traditional spoon-feeding approaches to introducing solids. A total of 206 women (41.3% primiparous) were recruited in late pregnancy from a single maternity hospital (response rate 23.4%) and randomized to Control (n = 101) or BLISS (n = 105) groups. All participants received government-funded Well Child care. BLISS participants also received support to exclusively breastfeed to 6 months and three educational sessions on BLISS (Baby-Led Weaning, modified to reduce the risk of iron deficiency, growth faltering, and choking) at 5.5, 7, and 9 months. Food variety was calculated from three-day weighed diet records at 7, 12, and 24 months. Questionnaires assessed infant preference for different tastes and textures at 12 months, and for ‘vegetables’, ‘fruit’, ‘meat and fish’, or ‘desserts’ at 24 months. At 24 months, 50.5% of participants provided diet record data, and 78.2% provided food preference data. BLISS participants had greater variety in ‘core’ (difference in counts over three days, 95% CI: 1.3, 0.4 to 2.2), ‘non-core’ (0.6, 0.2 to 0.9), and ‘meat and other protein’ (1.3, 0.8 to 1.9) foods at 7 months, and in ‘fruit and vegetable’ foods at 24 months (2, 0.4 to 3.6). The only differences in perceived food preferences observed were very small (i.e., <5% difference in score, at 12 months only). Infants following the modified Baby-Led Weaning were exposed to more varied and textured foods from an early age, but only an increased variety in ‘fruit and vegetable’ intake was apparent by two years of age.