<![CDATA[Google NotebookLM has attracted growing interest as a multimodal generative AI tool with potential for English as a Foreign Language instruction, yet its features have not been evaluated against established EFL pedagogical principles. This article presents a narrative literature review (NLR) of 50 articles to assess the extent to which NotebookLM's multimodal affordances correspond to those principles, and to identify the conditions under which its affordances weaken or fail. Five evaluative criteria were derived from three theoretical traditions: social semiotics and multiliteracies theory, the cognitive theory of multimedia learning, and CALL and task-based language teaching research. The analysis finds that the tool's source-grounded architecture meaningfully supports four affordances: audio transduction, cognitive load management under default conditions, pacing alignment, and authentic task affordance. Three further affordances remain unmet under the tool's current design: proficiency differentiation, full modal diversity, and contextual accessibility under constrained conditions. The review identifies three critical boundaries. Source-grounded generation carries architectural constraints and accuracy risk; frictionless compression suppresses Critical Framing; and native-speakerist audio defaults risk reinforcing hegemonic language norms. Across all findings, teacher mediation emerges as the constitutive condition for the actualisation of affordances. The findings suggest that NotebookLM functions most effectively as a teacher-dependent instrument for multimodal content engagement, and that its responsible deployment in EFL contexts requires deliberate source curation, scaffolded task design, and explicit critical AI literacy instruction.]]>
Copyrights © 2026
