<p>Neurofeedback (NFB) is a non-invasive intervention for ADHD that aims to train self-regulation of brain activity. Despite decades of clinical application, evidence for its efficacy remains mixed, and the mechanisms—particularly the role of NFB learning in ADHD symptom reduction—are still not well understood. This study advances the NFB field by introducing both theoretically grounded and data-driven frameworks that conceptualize learning as a dynamic, individualized process. Data were drawn from a 25-month longitudinal randomized controlled trial of theta/beta ratio (TBR) NFB treatment in children with ADHD (<i>N</i> = 100; ages 7–10). Moving beyond prior work that characterized learning through binary classifications and group-level averages, we captured within- and between-session training dynamics and applied clustering methods to define distinct learner profiles, highlighting the heterogeneous nature of NFB responses. Our results demonstrate that while most participants learned to regulate TBR effectively, individual learning did not reliably predict symptom improvement. Instead, linear mixed-effects model results indicated that non-specific factors such as participant engagement and training context were more predictive of clinical outcomes. Moreover, variability in learning curve shapes—typically overlooked in standard NFB studies—suggests that NFB learning does not follow a uniform pattern, with individuals engaging through distinct mechanisms aligned with either operant-conditioning or skill-acquisition models. These insights advance theoretical understanding of NFB mechanisms and call for individualized, mechanism-informed approaches to optimize treatment outcomes and deepen clinical interpretation.</p>