This research investigates the impact of an AI adaptive language assessment system, when combined with special education principles, on neurodiverse students in TESOL contexts. Although adaptive systems have been extensively debated in language learning, there has been remarkably little attention paid to students with autism spectrum disorder, dyslexia, or ADHD. To fill this research void, the study employed a sequential explanatory mixed-methods approach. In the quantitative component, 120 students were included in a 12-week quasi-experimental design comparing the impact of AI adaptive assessment with traditional testing modes. The data included standardized English proficiency test scores, test anxiety, engagement, and psychometric statistics using Item Response Theory and differential item functioning. The results demonstrated greater proficiency achievement, reduced anxiety, and increased engagement among students using the adaptive system. Reliability coefficients were high, and subgroup analysis revealed little measurement bias. In the qualitative component, teacher interviews shed light on usability and integration in the classroom. In general, the results of this study indicate that by combining adaptive assessment with organized special education principles, students with diverse cognitive abilities can be treated equitably and meaningfully in language assessment, while also offering a roadmap for future research on transparency and long-term implementation.