In this channel's material, inline citations function as the trust mechanic of a sources-grounded research workflow. The concept appears in the NotebookLM walkthrough, where every answer the model produces carries numbered markers showing exactly which source passage generated it [^0cc0dd07-dc88-5d83-8cf7-2f6e48821796@1032]. Citations show up twice in the demonstrated workflow. First, they are what the model returns: Johnson's reader guide prompt — context about himself plus a request for the most important sections to read — comes back as an interview-by-interview guide with citations [^0cc0dd07-dc88-5d83-8cf7-2f6e48821796@2141]. Second, they are what you act on: the click-through step, where clicking a relevant passage's citation saves it onto the noteboard via add to note [^0cc0dd07-dc88-5d83-8cf7-2f6e48821796@2204]. The practical takeaway for your own workflow: treat inline citations as your verification layer, not decoration. Instead of accepting an AI answer wholesale, use the citations as a reading scaffold — click through to the original passage, judge it yourself, and pin what resonates. That loop is what let the session move from 200,000 words of NASA oral-history transcripts to a curated set of notes and, eventually, an Apollo 1 documentary concept and opening script. To explore the technique in context, jump into the moments [^0cc0dd07-dc88-5d83-8cf7-2f6e48821796@52].