Google just expanded NotebookLM and Gemini in ways that make them far more useful for serious research and everyday automation. If you create notes, automate work, or build AI-driven workflows, these updates are worth learning now. Core improvements include persistent chat history, deep research that cites many sources, broader file type support, customizable video and audio overviews, and a robust no code space to build AI agents using Gemini and Flows.
Table of Contents
- Why these updates matter
- What changed in NotebookLM
- Gemini, Flows, and AI agents
- Practical examples and templates
- Activity logs and observability
- Security, privacy, and best practices
- How to get started step by step
- Suggested visuals and multimedia
- Internal and external linking suggestions
- Meta description and tags
- Call to action
- Frequently asked questions
- Final thoughts
Why these updates matter
NotebookLM is no longer just a quick Q and A playground. It now functions like a research assistant you can customize, store, and re-run. Meanwhile Gemini plus flows.workspace.google.com lets you convert ideas into automated agents that can watch email, summarize documents, post to chat, and connect to third party tools. Together these changes move Google from conversation prototypes to usable productivity tools that save time and reduce cognitive load.
What changed in NotebookLM
The NotebookLM updates can be grouped into five main improvements. Each makes the product more practical for real work.
- Persistent chat history so sessions remain when you close and reopen a notebook.
- Deep research mode that runs a multi step sourcing process across the web or your Drive and returns many cited sources.
- Expanded file support allowing Google Sheets, Drive URLs, images such as handwritten notes, PDFs from Drive, and Microsoft Word documents.
- Customizable video and audio overviews so you can control format, length, visuals, and target use case for generated overviews.
- Chat customization enabling you to set conversational goal, style, role, and default response length.
1. Persistent chat history
NotebookLM keeps your conversations inside each notebook so you can pick them up later. That alone changes how you work with it. Previously you had to manually save chat threads as notes to keep them. Now the session lives with the notebook until you choose to clear it.
Important note, refreshing the chat will clear history. Only refresh if you have saved any content you want to keep to a note. I still recommend saving important conversations as explicit notes to avoid accidental loss.
2. Deep research versus fast research
Fast research remains useful for quick lookups. Deep research is where NotebookLM shines now. When you select deep research and provide a topic, NotebookLM runs a multi step plan: it outlines a research approach, searches sources, synthesizes findings, and returns a richer set of citations and a structured output. In my tests, deep research cited dozens of sources rather than the handful returned by fast research.
Use deep research when you need a thorough primer, a literature style review, or to prepare for a talk or report. Configure it to search only your Drive if you want private, organization specific results.
3. New file types you can upload
NotebookLM now accepts a wider range of inputs. That matters because real work documents are rarely plain text. The new supported file types include:
- Google Sheets
- Google Drive files via URL
- Images such as photos of handwritten notes or brochures
- PDFs stored in Google Drive
- Microsoft Word documents
If your workflow relies on mixed media, you can now centralize research into a single notebook and ask NotebookLM targeted questions across those inputs.
4. Audio and video overview customization
NotebookLM now generates customizable audio and video overviews. Previously you chose between broad presets like deep dive or brief critique. Now you can add custom instructions for the overview. For video overviews you can control:
- Type of overview (explainer, brief, show structure)
- Visual style or ask it to auto select visuals
- Target use case such as an internal briefing or public explainer
- Which sources the overview should focus on
For audio overviews you can specify focus points and instruction for the narration. This is great when you need consistent, reusable content like meeting briefings or quick podcast style summaries.
5. Chat customization
You can now configure the chat persona and output format. Choose a conversational goal, role, and style. For example set the model to respond at a PhD level or to role play as a tabletop game host. You can also set the default response length. I prefer shorter responses for concise insights, but the flexibility is helpful if you need fuller explanations.
Gemini, Flows, and AI agents
The no code automation environment at flows.workspace.google.com allows you to create AI agents that automate tasks across Google Workspace and a growing list of external services.
There are two main approaches for building agents. Describe the task you want automated and let Gemini build the agent for you. Or create an agent manually using the flow builder, which lets you choose triggers, logic, and actions step by step.
Triggers and starters
The agent begins with a starter. Starters can be scheduled or event driven. Examples include:
- New email arrives
- New chat message
- Form response received
- Meeting starts
- File edited
- Item added to a folder
- Spreadsheet row added or changed
Combining events opens powerful patterns. For example, a sheet update can trigger a search across Drive, then generate a chat summary with action items.
Steps, actions, and tools
Each flow is built from modular steps. Key operations include:
- Ask Gemini to analyze or summarize
- Extract specific fields
- Recap unread emails
- Decision logic and filters
- Integrations like webhooks
- Actions on Google services: Gmail, Chat, Sheets, Drive, Calendar, Docs, Tasks
Google also includes early integrations with tools like Asana, Confluence, Jira, MailChimp, QuickBooks, and Salesforce. Expect that list to expand quickly.
Practical examples and templates
Here are practical agent templates you can adopt or customize immediately.
Daily unread email summary
- Starter: Scheduled every morning at 8 am
- Step 1: Recap all unread emails
- Step 2: Gemini summarizes key items and extracts actionable highlights
- Step 3: Notify you in Chat or send a condensed email
This is perfect for reducing inbox anxiety while ensuring critical messages are surfaced quickly.
Pre meeting brief
- Starter: Trigger when a meeting begins or 30 minutes before the meeting
- Step 1: Gather recent emails with the meeting subject, relevant Drive docs, and calendar notes
- Step 2: Ask Gemini to create a brief with talking points and anticipated questions
- Step 3: Send the brief to Chat or email attendees
This saves prep time and makes meetings far more productive.
Urgent email alerts
- Starter: New email arrives
- Step 1: Filter emails for keywords like refund, billing failed, or account status
- Step 2: If the filter matches, send an immediate Chat notification and add a task
I use urgent alerts for sponsor requests and failed automations. This prevents missed opportunities and noisy mornings.
Activity logs and observability
Flows provides an activity log with up to 40 days of history. Use it to monitor execution, debug failed steps, and confirm that decisions were applied correctly. The activity log shows statuses such as complete, in progress, and had issues. If a flow fails, the log is the first place to look to identify which step needs adjusting.
Security, privacy, and best practices
As you adopt these tools, follow these best practices to reduce risk.
- Save important chat content to a note before refreshing or clearing history.
- Limit agent permissions to only the data sources they need.
- Test agents in a development account before running them against production mailboxes or calendars.
- Review activity logs regularly so you can spot unexpected behavior early.
- Audit third party integrations to ensure they meet your organization security policies.
How to get started step by step
- Create or open a notebook in NotebookLM. Ask a simple question so you see chat history persist.
- Try deep research on a topic such as building AI agents for beginners. Compare results to fast research to understand the depth difference.
- Upload mixed files: a PDF, a Google Sheet, and a photo of handwritten notes to see how NotebookLM ingests them.
- Customize chat persona and default response length for your needs.
- Visit flows.workspace.google.com and explore the Discover templates. Pick a template like daily email summary and install it as a starting point.
- Modify triggers and steps. Add a filter for keywords you care about and test the agent on a sample email.
- Monitor the activity log for 48 hours and refine prompts, filters, and notifications to reduce noise.
Suggested visuals and multimedia
Including images and short videos will help readers adopt these features. Suggested media to add:
- Screenshot of deep research results with the citation count highlighted. Alt text: “NotebookLM deep research results showing 34 citations.”
- Screenshot of the NotebookLM file upload dialog listing Google Sheets, PDFs, images. Alt text: “NotebookLM upload options include Google Sheets, Drive URLs, PDFs, Word.”
- Flow builder screenshot showing starter and steps. Alt text: “Gemini flows builder with starter, steps, and integrations.”
- Short screencast of a pre meeting brief being generated and sent to chat. Alt text: “Screencast of automated pre meeting brief delivered in Google Chat.”
Internal and external linking suggestions
For internal navigation, link to pages that teach AI workflows and automation. Examples: an internal article about AI driven email workflows, a tutorial on designing prompts, and a catalog of agent templates you maintain. For external authority, link to Google’s NotebookLM documentation, Google Workspace developer docs for Flows and APIs, and reputable research about AI automation impacts such as industry reports estimating workforce changes.
Meta description and tags
Meta description (150 to 160 characters): Google upgraded NotebookLM and Gemini with persistent chat, deep research, new file support, and no code AI agents for automating work.
Suggested tags and categories: NotebookLM, Gemini, Google Flows, AI agents, AI automation, productivity, Google Workspace, deep research, prompt engineering.
Call to action
Start by testing deep research on a topic you care about and build one simple agent that saves you time each week. If you want step by step templates and an audit of your personal AI workflow, consider enrolling in structured training that covers prompt design, agent architecture, and privacy best practices.
Frequently asked questions
What is the difference between deep research and fast research in NotebookLM?
Fast research returns a quick set of sources and short summaries for rapid lookups. Deep research runs a multi step plan that outlines research goals, searches a broader set of sources, synthesizes findings, and provides many more citations and a structured result. Deep research is better for thorough learning or report preparation.
Can NotebookLM search my Google Drive only?
Yes. When you run research, you can choose to search the web or restrict the search to your Google Drive. Restricting to Drive is useful for private, company specific investigations and avoids public web noise.
Which file types can I upload to NotebookLM now?
NotebookLM now accepts Google Sheets, Google Drive files via URL, images such as photos of handwritten notes, PDFs stored in Drive, and Microsoft Word documents. These additions let you centralize mixed media research into one notebook.
How do Gemini agents get triggered?
Agents can be triggered on a schedule or by events. Available starters include new emails, chat messages, form responses, meetings, file edits, items added to folders, and spreadsheet changes. You can combine triggers and add filters to reduce noise.
What integrations are available for flows and agents?
Out of the box, flows support actions on Gmail, Chat, Sheets, Drive, Calendar, Docs, and Tasks. There are early alpha integrations with Asana, Confluence, Jira, MailChimp, QuickBooks, and Salesforce. Webhooks and third party connectors extend possibilities further.
How long is agent activity logged for?
Activity data is available for up to 40 days. The activity log helps you debug issues and confirm successful runs. It shows statuses such as complete, in progress, and had issues.
Will refreshing a NotebookLM chat delete my conversation?
Yes, refreshing clears the chat history for that notebook. To avoid losing important content save your conversation as a note before refreshing.
Final thoughts
These updates make NotebookLM and Gemini far more practical for real work. Persistent chat history and deep research transform NotebookLM into a reliable research companion. Expanded file support and customizable audio and video overviews let you create consistent outputs for meetings and briefs. Gemini plus Flows delivers a growing no code platform to automate recurring tasks across Google Workspace and external tools. Adopt one new agent and one deep research workflow this week to see immediate gains.
If you want guidance on turning these features into automated workflows that save time and reduce stress, consider a structured course or a personalized workflow audit to accelerate adoption and avoid common pitfalls.
