Optimizing L2 English Reading Speed and Comprehension Through Artificial Intelligence: A Psycholinguistic and Pedagogical Framework

Abstract

Artificial intelligence (AI) is proposed as a means of improving second-language (L2) English reading. The pedagogical problem, however, is not maximizing words per minute but helping readers allocate processing resources so that lexical access becomes more efficient without sacrificing syntactic integration, inference, monitoring, and transfer. This paper synthesizes research on L2 reading, evidence on fluency and comprehension interventions, studies of intelligent tutoring and generative AI, and scholarship on assessment and educational ethics. The synthesis supports a layered model in which reading performance emerges from interactions among lexical and orthographic quality, decoding, sentence and discourse integration, working memory, prior knowledge, strategic regulation, motivation, and text-purpose conditions. Speed and comprehension are related but dissociable outcomes; therefore, AI should optimize a conditional speed–accuracy profile rather than a reading-rate target. The paper proposes a teacher-governed AI architecture combining learner modeling, text selection, repeated and extensive reading, vocabulary recurrence, strategy coaching, uncertainty-aware feedback, and delayed transfer assessment. These mechanisms are recommendations derived from converging evidence, not claims that AI automatically produces durable L2 gains. Evaluation should include rate, accuracy, literal and inferential comprehension, vocabulary breadth and depth, immediate and delayed transfer, learner agency, fairness, and teacher workload. The paper concludes that AI is most defensible as a scaffold whose assistance fades as learners develop self-regulated reading, while teachers retain responsibility for interpretation, assessment, and equitable participation.