NotebookLM 2.0 introduces several major upgrades that shift it from a simple summary assistant into a highly capable productivity engine grounded in user-provided data.

  • Generate Editable Files Directly From Chat: Instead of static, uneditable exports (like static slide images), users can prompt the chat to generate fully functional, editable files—including PowerPoint decks, Excel spreadsheets, PDFs, Word documents, and interactive charts.
  • Upgraded Backend Power: The platform now runs on Gemini 3.5. It displays a step-by-step thinking process in the chat interface and assigns a secure cloud computer to each notebook, enabling the AI to execute code for deep research and complex data analysis.
  • Smart Edit Agent Capabilities: Instead of manually searching or adding sources, users can ask the chat agent what topics or sources are missing. The agent scans the context, identifies gaps, performs the web research, and allows importing those new sources with a single click.
  • Grounding in Personal Data: The tool is highly optimized for personal datasets. Users can feed it raw files like ad spend sheets, revenue reports, or workout trackers, and have the chat analyze trends, cross-reference campaign performance, or output clean visual tracking charts.
  • Strict Grounding Policies: Despite web-search and code-execution abilities, NotebookLM remains strictly grounded. It will not pull outside information without explicit user approval, keeping the focus entirely on the user’s uploaded knowledge base.