📔 Lesson 2.2: Web, YouTube & Audio Sources (and Google Drive)

The Link Family — Same Panel, Different Rules

Everything from last lesson landed in the Sources panel, and so does everything in this one. But the "link family" — web pages, YouTube, audio, Drive, and the Discover helper — behaves a little differently from a file you upload, and the difference always comes back to the same golden rule from Lesson 2.1: NotebookLM works from readable text.

A file on your computer either has text in it or it doesn't, and you can check. A link adds a twist: NotebookLM has to go fetch the content first, and how much text it can grab depends on how the page, video, or recording was built. A clean article imports beautifully. A video with good captions imports its whole spoken content. But a page that hides its text behind scripts, or a video with no captions at all, can come in partial — or nearly empty. None of this is mysterious once you know what to look for, and that's exactly what this lesson gives you: the ability to predict, before you add a link, roughly how well it'll come through.

Website URLs

To add a web page, choose Add source and pick the website/URL option (in the current interface you paste a link into a box). NotebookLM visits the page and pulls out the readable text — the article, the post, the documentation — and turns it into a source you can chat with and cite, just like a PDF.

🌐 What imports well

Ordinary text-forward pages are the sweet spot: news and magazine articles, blog posts, documentation, reference pages, most Wikipedia-style content. If the meaning of the page is in its words and those words are served as normal text, NotebookLM grabs them cleanly.

⚠️ What imports poorly (the honest limits)

Some pages fight back, and it helps to know which before you're surprised:

• Paywalled or login-gated pages. If you can't read the full article without subscribing or signing in, NotebookLM generally can't either — it may pull only the free preview, or nothing.
• JavaScript-heavy pages. Many modern sites build their content in your browser with scripts. If the real text only appears after code runs, a fetch may see a mostly-empty shell and import very little.
• Interactive apps, dashboards, and infinite feeds. A social feed, a map, a web app — there's often no stable "article text" to extract.
• Image-heavy or PDF-behind-a-link pages. Same rule as ever: text imports, pictures mostly don't.

YouTube Videos & the Transcript Rule

This is the one that delights people: you can drop a YouTube link into NotebookLM and chat with the video — ask what it covered, get a summary, pull out the key points, and jump to cited moments. A two-hour lecture becomes something you can interrogate in minutes.

▶️ How it actually works — and why captions decide everything

Here's the single most important fact about YouTube sources, and it explains every success and every disappointment you'll ever have with them: NotebookLM reads the video's transcript — its captions — not the pixels of the video. It's using the words that were spoken, as text, exactly like every other source.

So the quality of a YouTube source depends almost entirely on the quality of its transcript:

• A video with good captions (creator-provided, or clean auto-generated ones) imports its whole spoken content and works wonderfully.
• A video with no captions at all — or with captions disabled — may import poorly or not work as a source. There's simply little text to read.
• Auto-caption quality varies. Heavy accents, background music, cross-talk, or technical jargon can garble the transcript, and NotebookLM inherits those errors.
• Visual-only content is invisible. A silent screen-recording, an on-screen chart nobody reads aloud, a wordless demo — if it isn't spoken, it isn't in the transcript, so the notebook won't know about it.

NotebookLM doesn't "watch" a video — it reads what was said. Pick videos where the value is in the talking, and you'll never be disappointed.

Audio Files as Sources

NotebookLM also accepts audio files — mp3 and similar formats — as sources. This is quietly powerful: a recorded lecture, an interview, a meeting recording, a voice memo, a podcast episode you've saved as a file can all become grounded, searchable, citeable material. NotebookLM transcribes the audio — turns the spoken words into text — and then treats that transcript as the source, the same pattern you just saw with YouTube.

🎧 What that means in practice

• Clear speech transcribes well. A well-recorded talk or interview comes through cleanly and becomes genuinely useful.
• Audio quality matters. Noise, crosstalk, distance from the mic, and strong accents can reduce transcription accuracy — the notebook can only be as good as the transcript.
• It's spoken words only. Music, tone, and non-speech sound aren't "content" the notebook can reason about — it's the words that count.

Discover Sources & Google Drive

Two more ways to add sources round out the picture — one that finds material for you, and one that taps the library you already have in Google.

🔍 Discover sources

Sometimes you have a topic but not the sources yet. NotebookLM's Discover sources helper lets you describe what you're researching in plain language, and it searches the web and suggests relevant pages you can add with a click. It's a running start when you're beginning from scratch, or a way to round out a notebook with a few authoritative pages you hadn't found.

Treat its suggestions the way you'd treat any search results: you're the editor. Discover casts a wide net; not every result will be high-quality or on-topic. Skim, pick the good ones, and skip the rest — because, as the next lesson hammers home, a few great sources beat a pile of mediocre ones. Note too that once a discovered page is added, it's a normal web source and inherits every web limit above.

📁 Adding from Google Drive

Because NotebookLM is a Google product, it plugs straight into Google Drive. When you add a source, you can open a Drive picker and select Google Docs, Slides, and PDFs stored in your Drive — no downloading and re-uploading. For anyone who already keeps their reading and writing in Google, this is the fastest on-ramp there is.

Remember the rule from last lesson, though: even a Drive Doc is imported as a snapshot copy. It feels live because the original stays in Drive, but the notebook read it at import time. Edit the Drive file later and the notebook won't pick up the change on its own. At the time of writing, a Google Doc or Slides source offers a sync with Google Drive option you can click to refresh it; for other files, re-import to refresh the source.

🎯 Project: A Web Page + a YouTube Video

This project makes the payoff of link sources real — and teaches you the single most magical thing about NotebookLM: it can answer across several sources at once. You'll add a web page and a captioned YouTube video on the same topic, then ask a question that only makes sense if the notebook read both.

"Compare what the article and the video say about X."
"What does the video add that the article doesn't cover?"
"Summarize the topic using both sources, and cite each."

Watch the citations: a good cross-source answer will point
into BOTH sources. If every citation is from just one, the
other may have imported thin -> confirm it and fix it.

Best Practices for Link Sources

✅ Do's

• Prefer text-forward pages and well-captioned videos. They import cleanly; you'll feel the difference immediately.
• Confirm every link after it fetches. A source in the panel isn't proof the text came through — ask it something only it would know.
• Use Drive for Google-native material. Docs, Slides, and PDFs already in Drive are the fastest sources to add.

❌ Don'ts

• Don't expect paywalled or app-like pages to import. If you can't read it without logging in, neither can NotebookLM — paste the text instead.
• Don't add captionless videos and hope. No transcript, no source. Check the captions first.
• Don't dump every Discover result in. Curate — a few great sources beat a heap of weak ones (next lesson).

📓 Learning Journal

📝 Lesson Summary

🎉 What You've Accomplished

You can now bring in almost anything — the web, video, audio, and your whole Drive — and, crucially, you can predict how well each will import and fix it when it comes in thin. You even asked your first cross-source question, which is the beating heart of real research. Your notebooks are no longer limited to the files on your hard drive.

❓ Common Questions at This Stage

🔭 Looking Ahead

You now know every way to get material in. Next is the craft of choosing wisely: Lesson 2.3: Curating a Source Set — Limits, Quality & Organization. You'll learn why a few great sources beat a pile of mediocre ones, how to select and deselect sources to focus the chat, how to rename and organize, and how source limits work — the difference between a notebook that gives sharp answers and one that gives confident nonsense.

✅ Before the Next Lesson

• Finish the project: a web page and a captioned YouTube video added, with a cross-source question asked
• Notice if any of your sources feel off-topic or low-quality — you'll curate them next lesson
• Write your Learning Journal entry for this lesson

📚 Additional Resources

• notebooklm.google.com — add a link source
• NotebookLM Help Center — web, YouTube & audio sources
• NotebookLM on the Google blog — source updates