Notebook LM has outpaced its flashier cousins in real use because it sticks to the truth. That’s the big lesson from Jeff Su’s latest breakdown—and it matches what I’ve seen as a customer-obsessed operator. My take: Notebook LM wins when accuracy matters and time is tight. If you’re still using it like a search bar, you’re wasting its best tricks.
The Case for Grounded Answers
Jeff’s stance is simple and sharp: Notebook LM shines when you know where the answers live and need the tool to read, connect, and produce. He shows how it ingests PDFs, spreadsheets, slides, audio, even video, and then delivers grounded outputs you can use.
“Notebook LM is still the perfect tool when three things are true… you know which files have the answers; the sources are in different formats; and you need the AI to stick to what’s in the documents.” — Jeff Su
As someone who helps brands earn superfans, this approach hits home. Customers reward teams that are precise and fast. Guesswork doesn’t earn repeat love. Grounded beats clever, every time.
What Jeff Gets Right
Jeff treats LM like a production tool, not a parlor trick. The three-panel flow—sources, chat, studio—turns chaos into clarity. Load sources, interrogate them, then ship deliverables. That last step is the unlock.
- Reports: Dynamic formats based on your sources, so you stop staring at a blank page.
- Slide Decks: Great for shaping a narrative fast, even if edits are clunky after export.
- Infographics: Clean visuals from source material; better if you upload brand guidelines.
- Mind Maps: A bird’s-eye scan to pick the few ideas that matter.
- Data Tables: Trustworthy comparisons grounded in the files you provided.
He’s blunt about weak spots too. Slide decks export as images, so edits take workarounds. Audio overviews feel like fluff unless you’re multitasking. And for heavy “think” work, he prefers other models.
“Notebook LM’s biggest strength—high accuracy—is also its biggest limitation—low creativity.” — Jeff Su
Smart Constraints, Better Outcomes
Jeff’s workflow shows discipline. He caps “web + fast research” sources at three to force manual checks. He writes custom instructions for each notebook goal. He clears chat history to prevent drift, then saves key insights as notes—or even promotes them to sources so they feed studio outputs. That’s how you get repeatable wins.
“Fast research gives you a list of sources to review. Deep research writes a report for you.” — Jeff Su
He argues against using deep research inside LM when your own expertise can filter junk better—or when other tools think longer. I agree. Use the right brain for the right job. Let LM read. Let a long-form model ideate.
Where I’d Push You Further
If your goal is loyalty, use LM to make your team feel psychic to customers. That means building “living” notebooks that always reflect the latest truth.
- Create a Customer Signals notebook: support transcripts, NPS comments, sales notes, and product updates. Generate monthly “What changed and why it matters” reports.
- Upload your brand voice and style guide. Tell reports, decks, and infographics to follow it. Consistency builds trust.
- Use mind maps before every major campaign. If a node doesn’t tie to a customer outcome, cut it.
This turns LM from a helper into a heartbeat. When the team can scan, align, and ship in a day, fans feel it.
Counterpoints Worth Noting
Could LM handle wild brainstorming? Not well. That’s fine. Let creativity live where it thrives. Bring those ideas back into LM to test them against facts. Jeff even uses it for tax prep, health trends, and meeting notes—all high-stakes, zero-fluff zones. That’s the sweet spot.
Final Thought and Call to Action
Jeff Su makes a strong case: Notebook LM is the right tool for truth-heavy work. Treat it like a production studio and it will save you hours while raising your hit rate.
My challenge to you: pick one business area this week—customer insights, finance, or competitive intel. Load your sources, write a sharp custom instruction, and generate a report or deck. Then ask one question: Did this help a customer faster? If yes, keep going. If not, refine your sources and try again. Your superfans are waiting for better answers—not better noise.
