Very few early childhood interventions have observed sustained effects regarding television viewing and none have examined the mechanisms behind sustained intervention effects at long-term follow-ups. Thus, the aim of this study was to investigate potential mechanisms relating to the maintained intervention effect on television viewing at two long-term follow-ups in the Melbourne Infant Feeding Activity and Nutrition Trial (INFANT). INFANT was a cluster-randomised controlled trial. At the 2- and 3.5-year follow-ups, a total of 262 infant/mother pairs had complete information. Television viewing was assessed via a questionnaire at both follow-ups and six potential mediators were measured post-intervention (i.e. 15 months after baseline). Causal mediation analysis was conducted. At the 2- and 3.5-year follow-ups, the positive impacts of INFANT on maternal television viewing knowledge were maintained (B = 0.34 units; 95% confidence interval (CI95): 0.21, 0.48). An indirect effect of the intervention on reducing childrens television viewing time was observed at the 2- and 3.5-year follow-ups (B = -11.73 min/day; CI95: -22.26, -3.28 and B = -4.78 min/day; CI95: -9.48, -0.99, respectively) via improved maternal television viewing knowledge. Conclusion: The positive impacts of INFANT on maternal television viewing knowledge were maintained at both follow-ups, with better maternal knowledge associated with less television viewing time in their children. These results have implications for paediatricians and healthcare professionals as educating new parents early on regarding screen time may lead to the development of healthier screen time habits that are sustained through to the pre-school years.