A systematic review of AI in second language acquisition using the expanded SAMR model (2015–2024)

This systematic review assesses the extent to which AI applications enhance versus transform SLA tasks. Following PRISMA guidelines, we searched Web of Science and Scopus (2015–2024), screened and included 281 studies after full-text assessment. The included studies were coded designs with the expanded SAMR model (product-process criteria), then analyzed overall distributions, skill-specific patterns, and emergent design blueprints. Results show that Augmentation studies dominated the field, with 35% of the reviewed studies achieving Modification or Redefinition status. Generative AI (GenAI), emerging after 2022, enabled but did not guarantee the adoption of higher-level designs. Writing and speaking comprised 60% of the corpus and mostly remained at enhancement; reading and listening together accounted for 12%, with about half of reading and one-third of listening studies reaching Modification through personalized or multimodal redesign. Vocabulary demonstrated the highest transformation ratio. Based on transformation-level cases, this review synthesizes a GenAI-driven three-stage framework (pre-class diagnostics/adaptation; in-class dialogic co-construction; post-class data-driven extension) and offer practical guidance for educators and policymakers.