Child eating behaviours have consistently been linked to child weight status. Yet, changes in child eating behaviours during early obesity treatment are rarely evaluated. Psychometric evaluation of the Child Eating Behaviour Questionnaire (CEBQ) is common, but results are sample-dependent and included items may not capture the full range of the underlying traits. Rasch analysis can overcome these disadvantages. The aim of this paper was to assess child eating behaviours measured by the CEBQ after a 12-month obesity intervention applying the Rasch model for the validation of the CEBQ. The Rasch-based fit statistics were applied in children from two samples, Australian and Swedish (n = 1724). Changes in eating behaviours amongst children aged 4–6 years were examined in the More and Less RCT for obesity treatment (n = 177), which compared a parenting programme (with and without boosters) against standard treatment. Parents completed the CEBQ at four time points over 12-months. Linear mixed models were applied to estimate treatment effects on the CEBQ, refined according to Rasch, over time. We found that the validity of CEBQ was confirmed after removing 4 items (item fit statistics outside range 0.5–1.5). When the refined CEBQ was used in the assessment of the RCT, there were no differences in parental reports of changes in children's eating behaviours between the parenting programme and standard treatment (group-by-time interactions p > 0.05). However, in the total sample food approach behaviours decreased while fussy eating behaviours increased (p < 0.05). In conclusion, the refined CEBQ proved to be a valid tool for examining parent-reported child eating behaviours. Early obesity treatment may decrease eating behaviours associated with higher child weight. Future research should address the associations between changes in child weight status and eating behaviours.