Artificial Intelligence in the Diagnosis and Treatment of ADHD: Opportunities, Evidence, and Challenges

Attention-deficit/hyperactivity disorder (ADHD) is a common neurodevelopmental condition whose diagnosis still rests on clinical interviews and rating scales, and whose treatment is limited by partial response to medication, side effects, and unequal access to behavioural care. Artificial intelligence (AI), and machine learning (ML) in particular, has been proposed as a way to make diagnosis more objective and treatment more personalised and scalable. This review summarises the current state of that evidence. In diagnosis, ML models trained on neuroimaging, electroencephalography (EEG), and behavioural or sensor data frequently report high accuracy, but results vary widely across datasets, are rarely validated externally, and have not yet produced a tool that clinicians can rely on. In treatment, the strongest evidence concerns a prescription digital therapeutic delivered as a video game-like intervention, whereas computerised cognitive training and neurofeedback show limited effects on blinded measures of ADHD symptoms. Key barriers are non-representative training data, algorithmic bias, privacy, weak regulatory and reporting standards, and the risk that predictive accuracy is mistaken for clinical utility. We conclude that AI is best positioned as decision support within a clinician-led, hybrid model of care, and we outline the research needed to move from promising prototypes to validated practice.