Introduction: Constraint-induced movement therapy (CIMT) improves upper limb (UL) motor execution in unilateral cerebral palsy (uCP). As these children also show motor planning deficits, action-observation training (AOT) might be of additional value. Here, we investigated the combined effect of AOT to CIMT and identified factors influencing treatment response. Methods: A total of 44 children with uCP (mean 9 years 6 months, SD 1 year 10 months) participated in a 9-day camp wearing a splint for 6 h/day and were allocated to the CIMT + AOT (n = 22) and the CIMT + placebo group (n = 22). The CIMT + AOT group received 15 h of AOT (i.e. video-observation) and executed the observed tasks, whilst the CIMT + AOT group watched videos free of biological motion and executed the same tasks. The primary outcome measure was bimanual performance. Secondary outcomes included measures of body function and activity level assessed before (T1), after the intervention (T2), and at 6 months follow-up (T3). Influencing factors included behavioural and neurological characteristics. Results: Although no between-groups differences were found (p > 0.05; η2 = 0–16), the addition of AOT led to higher gains in children with initially poorer bimanual performance (p = 0.02; η2 = 0.14). Both groups improved in all outcome measures after the intervention and retained the gains at follow up (p < 0.01; η2 = 0.02–0.71). Poor sensory function resulted in larger improvements in the total group (p = 0.03; η2 = 0.25) and high amounts of mirror movements tended to result in a better response to the additional AOT training (p = 0.06; η2 = 0.18). Improvements were similar irrespective of the type of brain lesion or corticospinal tract wiring pattern. Conclusions: Adding AOT to CIMT, resulted in a better outcome for children with poor motor function and high amounts of mirror movements. CIMT with or without AOT seems to be more beneficial for children with poor sensory function.