Background <p>Cerebral palsy (CP) is a group of permanent neurodevelopmental disorders affecting movement and posture, often accompanied by sensory, perceptual, and cognitive impairments. Action Observation Therapy (AOT), based on mirror neuron activation, has emerged as a promising approach for motor learning and rehabilitation. This study aimed to evaluate the effects of AOT on brain activation and motor-behavioral outcomes in children with CP.</p> Methods <p>Participants aged 7–15 years who met the inclusion criteria were randomly assigned to either the AOT Group (<i>n</i> = 12) or the Control Group (<i>n</i> = 12). The AOT Group received balance training combined with action observation videos, whereas the Control Group received conventional balance training with non-motor videos and written instructions. Both groups participated in the same exercise program for 30&#xa0;min per day, five days per week, for three weeks, with progressively increasing difficulty. Pre- and post-treatment assessments included the Pediatric Balance Scale as the primary outcome measure, along with secondary outcomes of functional mobility (Timed Up and Go Test), trunk control (Trunk Impairment Scale), and gross motor function (Gross Motor Function Measure-88 sections D and E). Cortical hemodynamic responses during balance tasks on hard and soft surfaces were assessed using functional Near Infrared Spectroscopy (fNIRS) as the mechanistic outcome measure.</p> Results <p>Both groups showed significant improvements in balance performance, TUG scores, trunk control, and gross motor function after treatment (<i>p</i> &lt; 0.05), with no significant between-group differences (<i>p</i> &gt; 0.05). Exercise training and ground type significantly influenced cerebral oxygenation, with variations in oxyhemoglobin concentrations across brain regions depending on time, group, and video observation interactions (<i>p</i> &lt; 0.05). Oxyhemoglobin levels decreased in the AOT Group, whereas the Control Group showed increases or surface-dependent fluctuations (<i>p</i> &lt; 0.05).</p> Conclusions <p>Both AOT and conventional motor learning-based balance training improved balance and motor function in children with CP. However, AOT did not provide additional functional benefits. Exercise training and ground type significantly influenced cerebral oxygenation. Decreased oxyhemoglobin levels in the AOT Group and variable responses in the Control Group suggest modulation of cortical activation by sensorimotor context.</p> <p><i>Trial registration</i> ClinicalTrials.gov (NCT05995197).</p>