This paper looks at Google's NotebookLM, an AI-powered research assistant tool that can represent dense academic content in a range of output modes, like FAQs, timelines, study guides, and, most uniquely, as "Deep Dive" discussions. The discussions mimic a talk-show, where two AI-hosts unpack complex ideas from reading or audio texts, for example, a journal article or lecture, connecting them with related ideas. The talk-show format provides a relaxed, low-pressure way for learners to grasp key concepts. NotebookLM is especially helpful for those who may find academic reading overwhelming, as it transforms the reading process into a more enjoyable, multimodal experience. The paper describes and evaluates the discussion generation function in relation to principles of second language and digital learning. Overall, NotebookLM shows promise in helping learners to understand and engage with academic content more deeply and may be especially valuable for second language learners in English Medium Instruction (EMI) or bilingual education contexts which emphasize content development alongside language learning. However, the tendency of the AI-hosts to extract, embellish and augment the content may lead to misrepresentation of the original source text, raising epistemological concerns about the spread of misinformation.