Generative AI as a cognitive assistive technology for adults with dyslexia: the role of anxiety, working memory, and metacognitive regulation.

PURPOSE: This study examined generative artificial intelligence (AI) as an emerging cognitive assistive technology for adults with dyslexia. Specifically, it investigated whether anxiety, working memory, and metacognitive regulation predict adaptive reliance on AI during problem-solving over time. MATERIALS AND METHODS: A short-term, two-wave repeated-measures observational design with a four-week interval was employed (N = 547 adults with dyslexia in Arabia). Participants completed standardised measures of anxiety, working memory, and metacognitive strategy use. Adaptive reliance on generative AI was assessed through transcript-coded AI-assisted problem-solving tasks. Cross-sectional structural equation models (SEM) were estimated at each wave, followed by a latent change score model (LCSM) to examine predictors of change. Bootstrap inference (500 resamples) was used to test indirect and total effects. RESULTS: Group-level means remained stable across four weeks; however, consistent multivariate patterns emerged. Anxiety was negatively associated with working memory and adaptive AI reliance at both waves. Longitudinally, baseline anxiety predicted declines in working memory and metacognitive regulation, as well as a small but significant decrease in adaptive reliance on AI. Improvements in metacognitive regulation were positively associated with increases in reliance on adaptive AI. Indirect effects were not supported, but the total longitudinal effect of anxiety on change in AI reliance was significant. CONCLUSIONS: Generative AI shows promise as a cognitive assistive technology for adults with dyslexia; however, affective and executive-control factors influence engagement with AI support. Anxiety appears to reduce reliance on AI for reflection during problem-solving. Interventions integrating anxiety-sensitive supports and explicit metacognitive scaffolding may enhance AI-assisted rehabilitation.