Hearing assessment in infants is complicated due to the lack of reliable behavioral responses. The BAMBINO (Behavioural Audiometry Measures in Babies: Innovation, Novelty and Optimisation) project explores the feasibility of automating hearing assessment in infants by leveraging facial behaviour analysis. This feasibility study investigates the potential of action units, head pose, and gaze direction to identify changes in behavioural response to suprathreshold sound stimuli in infants aged 7 to 24 months. Video recordings of 58 healthy infants were analysed using convolutional neural networks (1D-CNN and 2D-CNN), evaluated against human observers. Results indicate that head pose is the primary feature for classification, closely aligning with current clinical protocols, but facial expressions still provide additional insights. SignalGrad-CAM, was employed to interpret model decisions, revealing nuanced patterns such as subtle micro-expressions and subtle facial reactions that often preceded head turns. Post-hoc analyses regarding age and trial progression in the test set highlighted that infants aged 12–18 months were most responsive and performance declined in later trial stages, likely due to fatigue, highlighting the importance of selecting a proper session duration. Future work will address biases in the data, and subsequent phases of the BAMBINO project aim to extend this study to younger infants (3–7 months), including hearing-impaired populations.