We Need to Talk About NotebookLM
Is it the missing link in your AI game?
At the end of 2024, when Thought Leaders and Insightful Brands flooded our LinkedIn feeds with 2025 predictions, 2024 Best-Ofs, 2024 Worst-Ofs, and AI-generated LinkedIn Rewinds (which were brilliant, IMO), I was a keystroke away from joining in. I had a draft all ready to roll, and couldn’t wait to regale you, dear Readers, with insights from the first year of The Trending Communicator.
But I stayed my hand. Not because the insights and predictions were stale or dull (although I admit to some snooze-worthy ideas), but because I heard a little voice in my head—let’s call it my inner Mark Schaefer—telling me to be bold. Crazy. Audacious.
I won’t claim that what I ended up doing was any of those things, but it was different. Possibly unique (fact-check me on that, please). In case you missed it, I orchestrated an AI takeover of The Trending Communicator for one episode, wherein the AI hosts discussed some trends shaping the communications profession.
In their 15-minute repartee, they covered:
Shifting job descriptions in the evolving communication landscape
Leveraging AI as a tool to enhance human creativity and efficiency
Building a strong personal brand through authenticity and storytelling
Embracing continuous learning to stay ahead in a rapidly changing field
Honing forensic listening skills to navigate information overload
Integrating PESO Model© for a well-rounded communication strategy
Developing "T-shaped" skills to thrive in future communication roles
Kudos to them. But before I sing any more praises for my AI pod-jackers, let’s remember that the reason they did such a good job is that their source content is 20-something hours of human, long-form, unscripted, original conversations with communications and marketing leaders. In other words, they were good because their human input was good.
But even when the original content is stellar, it takes a combination of AI tools and models to make any derivative or repurposed content work. That’s why we have to talk about Google NotebookLM, the star of the show and the core of my content engine.
A NotebookLM Primer
If you don’t know anything about NotebookLM or know about it but aren’t familiar with it, think of it this way: NotebookLM is a generative AI tool that allows you to create personalized, customizable, private chatbots that use only the sources you provide. It’s available to anyone with a personal Google account. If you pay for Gemini Advanced or are on a Google Workspace Business Standard with a Gemini account, NotebookLM Plus is included if your Workspace administrator has deemed it so. Unless stated otherwise, everything I will share here is available in the free version. Here’s a basic overview of what you can do with it1:
Create your own niche mini-LLM (or SLM - small language model) for anything, any topic you can dream of. In the free version, you can create up to 100 separate Notebooks—that means 100 mini-LLMs. In the plus version, it’s 500.
Add up to 50 sources per Notebook, and each source can be up to the equivalent of 500,000 words or 200MB of information. So, each of your mini-LLMs can be fueled by up to 25 million words in total. That’s a lot of words. In the plus version, it’s 300 sources per Notebook. At the time of this writing, sources can be in the form of Google Docs, Google Slides, PDFs, .txt files, Markdown files, web pages (add the URL, and NotebookLM grabs the text from that page), public YouTube URLs, audio files, and cut-and-pasted text.
Use the chat function within the Notebook to deeply analyze the sources you provide and nothing more (it doesn't look for answers beyond the sources in the Notebook).
Summarize, synthesize, search, and analyze your sources, but don’t ask it to write good prose. You can ask it to write a blog post but it's not really made for that.
Save chat results you like as Notes, which can then be used as additional sources if you so desire (unlike the LLMs you’re used to, NotebookLM does not save your chat sessions, so you can’t return to a previous chat).
Generate an “Audio Deep Dive,” a conversation in which two AI voices discuss themes and topics drawn from the Notebook’s sources, with the click of a button. You can customize these to a point.
Share Notebooks for collaboration (Workspace users are limited to sharing within their organization)
Use generative AI in a secure and private environment. Google doesn’t use any of your Notebook content for training AI, and your information remains private unless you share it or choose to provide feedback to Google. At least, that’s what they say.
Now that you’ve got an idea of what you can do, you might be thinking, why? Why do I need NotebookLM? Why should I use it? I’ll give you seven reasons, all conveniently starting with R:
Researching: Create a Notebook on a topic you’re interested in or want to learn about or drop in sources you wish to analyze for a project (a book you want to write, for example). Then, dig into the sources with the chatbot to find out what you need to know.
Rethinking: Ever wonder if your past proclamations still make sense? Load up a Notebook with your older content and more recent sources that may or may not support your ideas. Ask the chatbot to help you review and evaluate your points of view and suggest where you might want to make some adjustments.
Rediscovering: Add as much of your original content as possible to a Notebook: podcasts, videos, blog posts, books, whatever you’ve got. Then go exploring. Ask the chatbot to create a briefing document about your content. Tell it to look for ideas you’ve had about topics or themes you want to explore. You’ll find insights from the past and maybe new interpretations you hadn’t considered.
Reviving: Like rediscovering, use your Notebook full of your older content to breathe new life into old words. Hunt through your older work and identify content that can be relevant again.
Remembering: Are you a prolific creator with a body of work so large you can’t possibly remember everything you’ve written or said? A corporate communicator who’s expected to recall what the CEO said in the 2015 annual report? Instead of combing through documents, create a Notebook and throw them all in. Then, ask the chatbot to find what you’re looking for and do your remembering for you.
Repurposing: Take your existing content and transform it into other formats or forms of media. Add your podcasts, white papers, books, technical reports, and/or video scripts into a Notebook and use the chatbot to extract insights and quotes for blog posts, social posts, and more. Combine this with researching, rethinking, rediscovering, refining, and remembering, and you’ll have a powerful content engine at your fingertips.
Reimagining: Or just straight-up imagining. You know those ideas you get in the shower, the ones that start with, “Wouldn’t it be great if”? Run to NotebookLM with your crazy idea and start a new Notebook. Load it with relevant content, get a little woo-woo, and let your mind connect new dots that haven’t appeared to you before. Experiment and play. Chat with the Notebook, ask outlandish or outrageous questions, and see how you start to think of the topic or content in new ways.
I can think of other reasons, and so can you, but I tapped out at seven Rs, so let’s take the win and move on.
Curiosity Unleashed or Gone Awry: A Totally Real Example
OK, I hear you, let me bring you back to ground level and make it all make sense. I wanted to write about NotebookLM because many of you have asked me how I did the NotebookLM takeover of my podcast. Now that you (hopefully) have grasped the basics of NotebookLM, at least in theory (plug time: contact me if you want a demo or tutorial), I am more than happy to pull back the curtain and share my wares. Here’s how it all went down:
The Idea: It started in the shower (of course) when I was struck by a wave of Reimagining. “Wouldn’t it be great if, instead of writing a 2024 roundup or 2025 predictions, I could have AI do it for me? But not writing…what if I could do it as an Audio Deep Dive...that I would then publish…as a PODCAST??”
The Execution: Once I knew I could make it work, I started my workflow with a visit to NotebookLM for some Researching, then went through my usual podcast publishing process with some minor adjustments.
Podcast Audio:
In a pique of Reimagining and Repurposing, I generated a customized Audio Deep Dive with the instructions, “You are taking over The Trending Communicator podcast. Talk about the most important trends facing the communications profession, according to guests of the podcast.” I went through four iterations (create the audio, listen, delete, create again) before I got the results I had hoped for.
I downloaded the .wav audio file and dropped it into Adobe Audition, added my intro and music.
I realized that it would only be proper for the music to also be AI-generated. I headed over to Suno.ai and created a techno-rock theme intro for a podcast; remarkably, I liked the first song it generated (normally, a visit to Suno is an automatic time loss of indeterminate duration). I downloaded the song and added it to my Audition workspace. Then I mixed down the audio and had my podcast content ready.
Episode Artwork:
All of my episode graphics feature the guest, so I wanted a photorealistic image of the NotebookLM AI “hosts” of the Audio Deep Dive. I had a description in mind and tried variations of “futuristic middle-aged man and woman of indeterminate age, human-looking, augmented with AI, dressed in sophisticated and modern black clothes” on Midjourney, Dall-E (ChatGPT), and Flux AI. If I told you I spent an hour, I might be lying because I lost track of time. But in the end, I had a few images to choose from and landed on one of the Flux creations2, which you can see in the image above.
I dropped the image into my episode graphic template in Canva, found a font that looked creepy and tech-forward, finalized the design, and voila! Artwork done.
Show Notes:
I decided long ago (well, last year) to use AI for my show notes, but I always write the first few paragraphs myself. This time, I went 100% all-in on AI and let the chips fall as they might.
I uploaded the completed audio file into two tools I love: Flowsend.ai and Fireflies.ai.
I used Flowsend to generate the episode title, main show notes, notable quotes, and a high-level timestamped summary and copy-pasted it into Libsyn, my podcast host and publishing platform. I corrected a few spelling errors but what you see on the episode page is virtually unchanged.
In Fireflies, I generated the transcript and episode highlights (I have a soft spot for Fireflies’ emoji bullets). I downloaded the transcript as a .SRT file and added that to the episode assets on Libsyn, and copy-pasted the highlights as above.
Once all of these pieces were in place, I published the episode to all podcast platforms via Libsyn. Once published, Libsyn also exports the podcast to my Podpage podcast website. Podpage automatically posts to Facebook and Twitter/X.
LinkedIn Post and Audiogram:
Someday, I’ll get into video, but for now, I create audiograms to accompany my LinkedIn posts.
I cut a 30-second clip from the audio file and dropped it into Wavve.co, a video creation tool I’ve been using for years. In Wavve, I used the episode graphic as a background image, chose a waveform style, added captions, and generated the mp4 audiogram.
I went back to Flowsend, grabbed the LinkedIn post it had generated, and copied it into a new LinkedIn post. I added the audiogram to the post, made a few adjustments to the copy, and when I felt I was good and ready, hit the button. Here’s the final audiogram, in case you want to throw 54 seconds of your life away:
Give Me More of the New Stuff
I admit there’s a shiny-new-toy vibe to NotebookLM, and I do tend toward exuberance when I get to play with new toys. Anyone listening at my office door would have heard the occasional cackle, chortle, and giggle while I was bringing the AI podcast takeover concept to life. I wouldn’t blame you for questioning my idea of fun (and sanity). Still, the result is a solid example of what could happen when you let AI-savvy, technophilic, and creative storytellers follow their curiosity: fresh, reimagined, or revived content created quickly, efficiently, and cost-effectively and produced end-to-end by one person.
If you read through the play-by-play above, you might doubt my claims of speed and efficiency. Could my workflow have been more efficient? Without a doubt, and if I ever do this again, I’ll get it done in half the time or less. But even so, five minutes ago I’d have needed a writer, a graphic designer, and an A/V specialist (or two) just to create something like this. If this were a project for a client or an employer, I wouldn’t be playing around (ahem, Midjourney and Suno) as much and I would be more directed and focused in my prompting. I’d start with a clear brief, take inventory of my creative assets, map out my workflow, decide which AI tools and software would best fit the task, and knock this out in no time. And each time I do it, I’d find more efficiencies, refine templates and processes, and create quality content in less time.
Takeaways
Let me try to land this plane. Just in case I haven’t made it clear that I’m enamored with NotebookLM and can hardly contain myself, here are a few thoughts to close out this love letter:
During World War II, Alan Turing and Gordon Welchman invented the Bombe, a machine designed to break the German Enigma codes. The Bombe machine and Turing’s work were arguably the most critical developments in the history of military intelligence and played a significant role in winning the war. To me, NotebookLM is the Bombe machine, decoding the Enigmas of content production. Hyperbole? Me? Never.
With the right questions and prompts, NotebookLM alone reveals insights and answers that either would have remained hidden or taken untold hours of research and review to find.
If you’re the curious and creative type, NotebookLM is your imagination’s best friend. You’re only limited by the sources available to you and Google’s massive context window (the amount of data you can use).
If you can access various LLMs—ChatGPT, Claude, Gemini, etc.—using them in conjunction with NotebookLM is a creative multiplier. For example, cut and paste Notes you’ve saved into Claude and transform them into a speech or article. Paste multiple notes into a Google Doc or Word doc and use that as a source for a CustomGPT or project knowledge for Claude. Endless possibilities.
Setting up Notebooks within NotebookLM is simple and requires no technical skill. The availability of sources is your only obstacle to setting up discrete Notebooks for any topic, theme, subject, author, or crazy idea you’re interested in.
Even though NotebookLM could be the missing ingredient for your special content sauce, it’s only one ingredient. What makes the sauce special is the creative mind behind the recipe (that’s you). Remember, your uniquely human qualities and advantages - curiosity, creativity, craziness, critical thinking, and the ability to take showers - determine the outcomes of your interplay with NotebookLM or any other generative AI tool.
Finally, I want to help you understand NotebookLM not only because it’s Just That Important, but because I am convinced that it will unlock creativity, accelerate innovation, improve the quality and depth of our content, and help us all stay relevant. Let me know what you think in the comments below. And if you want my help making NotebookLM work for you or your organization, let’s talk.
Notes
This post is almost 100% human-written content. Almost.
The bullet points in the opening section were cut-and-pasted from AI-generated text. Apart from that, I wrote every word.
Grammarly gave me a helpful nudge here and there.
I used Midjourney to generate the hero image.
Are you using NotebookLM? Do you want to learn more? Drop a comment below or hit me up on LinkedIn. And don't forget to subscribe to The Trending Communicator podcast for more insights on navigating this wild, AI-powered world.
I’d be happy to show you how to do all of this. I offer workshops, webinars, Q&A sessions, and customized training on NotebookLM and generative AI. Contact me to hear more.
If you’re interested, here’s the optimized Flux.io prompt: “Against a sleek black backdrop, a cybernetically enhanced middle-aged man and woman exude confidence, their charismatic smiles captivating the viewer. Infused with advanced AI, their human features are subtly augmented, while their futuristic all-black corporate chic attire pulses with illuminated circuits that weave through the fabric. As they move, dynamic lighting cascades over their faces, creating an elegant glow that accentuates their modern sophistication. The scene masterfully blends power and style, inviting viewers into a mesmerizing world where cutting-edge technology and high fashion converge in a captivating dance of light and motion.”
Dan, you are my AI muse! I can't tell you enough how helpful and informative I find your posts. Huge thanks from one of your loyal readers