Thing 7

AI notebooks and knowledge management

Last reviewed: March 2026 30–45 minutes

In Thing 6, you uploaded a document to a chatbot and asked it questions. That's useful, but each conversation starts fresh; the AI doesn't remember what you uploaded before. AI notebooks let you build a lasting collection of sources and keep coming back to them, more like giving a colleague a whole project folder than asking them to read a single report.

AI notebooks let you gather PDFs, web pages, YouTube videos, and other sources into one workspace, ask questions that draw on everything at once, and come back weeks later to pick up where you left off.

The standout tool in this space is Google NotebookLM. Beyond the usual document chat, it can take your sources and turn them into a podcast-style audio conversation: two AI voices discuss your material, draw connections between documents, and explain the main points. It's an odd experience the first time you hear it, and a surprisingly effective way to absorb information you haven't had time to read.

This Thing is about moving from one-off document queries to something closer to a personal research assistant, one that knows your sources, stays grounded in them, and can present what it finds in ways you might not expect.


What makes AI notebooks different

An illustration representing AI notebooks and knowledge management, with interconnected documents and sources
AI notebooks let you build persistent collections of sources that stay available across sessions.

If you've been using ChatGPT, Claude, or Gemini for the previous Things, you've been working within individual conversations. Each conversation starts fresh. You upload a document, ask your questions, get your answers, and when you close the tab or start a new chat, the context is gone. Some chatbots now offer memory features that carry certain details between conversations, but they don't retain the actual documents you uploaded or the full context of your analysis.

AI notebooks take a different approach. Instead of conversations, you create a notebook, a workspace built around a specific topic or project. You add sources to that notebook, and they stay there. Every time you open it, all your sources are available. When you ask a question, the AI draws on everything in the notebook to answer, and it tells you exactly which source each part of its answer came from.

The difference is worth noticing. In a chatbot conversation, you're saying "read this and help me." In a notebook, you're saying "here's everything I'm working with; help me make sense of it all."

The grounding principle from Thing 6 still applies, and it's even stronger here. NotebookLM deliberately restricts its answers to your uploaded sources. If you ask a question that your sources don't cover, it will tell you rather than filling in from its general training data. This makes it unusually reliable for research and analysis, because you can trust that what it tells you actually came from your material.


Google NotebookLM: the main tool

Google NotebookLM is free with a Google account and is currently the most capable AI notebook tool available.


Beyond NotebookLM: other options

NotebookLM is the most fully featured AI notebook tool available for free, but it's not the only approach.


What to upload and when to use notebooks

The same privacy considerations from Thing 6 apply here, and arguably matter even more. With a notebook, you're building a persistent collection of sources that lives on Google's servers (for NotebookLM) or the relevant provider's infrastructure. That's fine for publicly available documents, content you've created yourself, or personal study materials. Think carefully about anything containing sensitive or confidential information.

Privacy reminder: for this programme's activity, work with sources you're comfortable uploading: publicly available reports, articles, YouTube videos, or materials you've created yourself. Don't upload anything from your workplace that might be confidential or covered by your employer's data policies.

AI notebooks are particularly useful when you're working with multiple sources over time: research projects, professional development, preparing for a complex meeting, or building expertise in a specific area. The persistence and source grounding make them a solid long-term reference tool.

They're less necessary for quick, one-off tasks. If you just need to summarise a single PDF or ask a few questions about one document, uploading it to a chatbot conversation (as you did in Thing 6) is faster and simpler.

A good rule of thumb: if you'll want to come back to these sources again, or if you're working with more than two or three documents on the same topic, a notebook is worth the small extra effort. For a quick query about a single document, a chatbot conversation will do.


Resources to explore

Google NotebookLM

Free with a Google account. The primary tool for this Thing. Supports PDFs, web URLs, YouTube videos, Google Docs, and more.

Visit tool
OpenAI Notebook

OpenAI's document-grounded notebook tool within the ChatGPT ecosystem. Newer and less feature-rich than NotebookLM, but worth knowing about.

Visit tool
Claude Projects

The Projects feature allows persistent document collections across conversations. Free tier available with some limitations.

Visit tool
NotebookLM help centre

Google's official documentation, useful if you want to explore advanced features or troubleshoot.

Read docs

Activity: build a notebook, hear your sources

30–45 minutes Google NotebookLM (free Google account required)

This activity has two parts: building a notebook and generating an audio overview. You'll end up with something you can actually listen to.

Part 1: build your notebook

  1. Create a notebook. Go to notebooklm.google.com, sign in with a Google account, and create a new notebook. Give it a name related to a topic that genuinely interests you: a professional interest, a hobby, or something you've been meaning to research.
  2. Add three to five sources. Aim for a mix of source types if you can: a PDF report, a couple of web articles, a YouTube video. Good places to find sources:
    • Government or charity reports in your area of interest (many are freely available as PDFs)
    • Articles from reputable publications
    • YouTube talks, lectures, or explainer videos with decent transcripts
    • Wikipedia articles (NotebookLM can import web pages)
  3. Chat with your notebook. Spend some time asking questions. Try:
    • "What are the main themes across all of these sources?"
    • "Where do my sources agree, and where do they differ?"
    • "Summarise what source [X] says about [topic] and compare it with source [Y]."
    • "What questions do these sources leave unanswered?"
  4. Check the citations. Click through to a few citations in the responses to verify that the AI is representing your sources accurately.

Part 2: generate an audio overview

  1. Generate an Audio Overview. Before you hit generate, customise the focus. Try giving it a specific angle, like "focus on the practical implications" or "explain this as if the listener is new to this topic."
  2. Listen to the full overview. It will likely run between five and fifteen minutes. As you listen, consider:
    • Does it accurately represent what's in your sources?
    • Does it surface any connections or themes you hadn't noticed?
    • How does the experience of listening compare to reading the same material?
    • Would you use this as a way to review material in future?

Your output

A document or blog post containing:

  • The topic you chose and a list of the sources you added (with links where available)
  • Three to five of the questions you asked in the chat, along with the AI's responses (copied or screenshotted), with a note on whether the citations checked out
  • The audio overview file (you can download it from NotebookLM) or a screenshot of the audio overview screen
  • A short reflection (a few paragraphs) covering: how did the notebook experience compare to the single-document analysis in Thing 6? Was the audio overview useful, gimmicky, or somewhere in between? Can you see yourself using an AI notebook for ongoing work or study?
Privacy reminder: use personal interests, publicly available sources, or content you've created yourself. Never upload confidential work materials.

Why this matters

Professionals rarely make decisions based on a single document. You read multiple reports, compare perspectives, synthesise findings, and build understanding over time. AI notebooks support exactly this kind of work. The audio overview offers a way to engage with material that most people haven't tried before.

Even if you decide notebooks aren't for you, the audio overview alone is worth trying.


Claim your Open Badge

Submit your source list, chat examples with citation checks, evidence of the audio overview (the file or a screenshot), and your reflective commentary as evidence for your Thing 7 badge via cred.scot.

Thing 7: AI notebooks and knowledge management open badge
Thing 7: AI notebooks and knowledge management

Submit your notebook activity output as evidence to claim this badge via cred.scot.

Claim now

What's next

In Thing 8, we'll look at AI-powered deep research. Where NotebookLM works with sources you've already found, deep research tools go out and find sources for you: they browse the web, read dozens of pages, and produce structured research reports with citations. If this Thing gave you a personal research assistant, Thing 8 gives that assistant the ability to visit the library on your behalf.