NotebookLM now integrates Deep Research to automate complex searches and adds support for more file types you use every day. Can you imagine delegating the heavy lifting of search and keeping to build your notebook while the system fetches sources and an organized report? That's exactly what they present.
Deep Research: your research agent inside the notebook
Deep Research acts like a dedicated researcher: it takes your question, designs a search plan, and explores hundreds of websites, refining queries as it learns. In minutes it gives you an organized report, with citations and links, ready to add to your notebook.
What makes it different? First, the report isn't a separate PDF: you can import the report and its sources directly into your notebook and keep adding material while the agent runs in the background. Second, you can then use NotebookLM's features—like audio or video summaries—to transform those findings and understand them better.
How it works (technically, but accessible)
Deep Researchcreates a search plan: it defines keywords, target sources, and quality criteria.- It runs multiple queries and controlled scrapes, extracts relevant snippets, saves metadata (URL, author, date) and evaluates basic source reliability.
- It consolidates the information into a synthesized and source-grounded report (that is, with traceable references) which you can review, import, and link into your workflow.
Want more control? You can direct the agent to search specific domains or exclude sources. That's useful when you need reproducibility or when you want to prioritize academic literature over blog notes.
How to try Deep Research in NotebookLM
- Open the sources panel and select Web as the source.
- Define your query and choose the research style:
- Fast Research: a quick scan to get immediate results and review sources fast.
- Deep Research: a more exhaustive analysis that searches, refines and synthesizes findings into a full report.
- Let the agent run in the background while you add other sources or work in your notebook.
Support for more file types
NotebookLM expands source input to cover the real formats you work with. That makes it easier to keep everything in one place without constantly downloading and re-uploading files.
- Google Sheets: you can now add sheets and ask about key statistics, generate summaries or request insights from structured tables.
- Drive files as URLs: paste Drive links as if they were web pages or videos. You can add multiple links separated by commas to import several files at once.
- Images: upload photos—like handwritten notes or a flyer—so the system extracts text and includes that content in the research.
- PDFs from Google Drive: add articles, reports or books directly from Drive without downloading and re-uploading.
- Microsoft Word (
.docx): upload drafts, notes or documents for analysis and idea extraction.
These features will roll out to all users within the week after publication, with image support expanding in the following weeks.
Technical tips and considerations for advanced users
- If you work with sensitive data, review NotebookLM's usage and privacy policies before importing confidential material.
- For reproducible analysis, specify preferred domains and save imported URLs:
Deep Researchstores metadata that makes auditing easier. - In Google Sheets, structuring columns well (dates, identifiers, metrics) improves the quality of automatic summaries.
- Keep source verification in mind: the agent helps evaluate basic reliability, but critical checks (cross-checking studies, reviewing methods) remain the researcher's responsibility.
Why this matters for you
The novelty isn't just more supported formats. It's that you can now combine an agent that searches at scale with your own workflow in a notebook that preserves context and traceability. Looking to write a paper, prepare a class, or validate hypotheses for a startup? You'll save hours and end up with a source-ready base to analyze and communicate.
Original source
https://blog.google/technology/google-labs/notebooklm-deep-research-file-types