Prompted: We Tested the Viral AI Podcast Generator …

You’ve probably seen the headlines: “Generate a Podcast in Seconds!” “A New Hit Podcast Tool That Will Blow Your Mind!” But is it just the digital marketing world hyping Google’s NotebookLM, or can this AI conversation generator really elevate your integrated marketing strategy? We took a deep dive (pun intended) and tested it on our own content to determine whether NotebookLM is worth incorporating into a martech stack. 

A Smarter Way to Use NotebookLM? 

Google positions NotebookLM as an “AI Research Assistant,” synthesizing and providing insights based on content a user inputs to the program. As a result, it’s meant to provide an accessible and general overview — not a customized content output. (We’ll talk about that more later.)

Powered by Google Gemini, the tool works by digesting your content inputs, or “notes,” and letting you interact with them. NotebookLM lets you engage with those notes in a few different ways, whether it’s conversations with its chatbot, generating study guides and FAQs, or creating what it calls a Deep Dive.

The Deep Dive is what has caught marketer’s attention. Based on the input notes, NotebookLM outputs a completely automated podcast-style conversation, with two AI hosts having a casual conversation back and forth about the information from your sources.

The Instant Podcast Experience

NotebookLM can digest up to 50 sources, including files from your Google Drive, website URLs, YouTube videos, and pasted text. We first input links for our monthly trend report, What’s Next, Now, but because the report is hosted on Substack, NotebookLM wasn’t able to crawl the page. So instead we fed it two posts from our Prompted blog series focused on the new world of AI search: How AI Is Changing the SEO Game and GEO vs. SEO: How to Optimize Your Content for Generative AI.

Originally the Deep Dive feature was completely automated, but Google has caught on to marketers’ interest and now lets users input brief text instructions on what the podcast should focus on. We gave it three prompts: What’s new in search, why it matters, and what you should do about it.

The Third Deep Dive’s the Charm

Each attempt to generate a new podcast takes a few minutes. While Google did introduce a new feature to tweak the direction of the generated podcast conversation, the tool still lacks the ability to make specific refinements. There’s also no option to edit or refine the length of the audio output.

Our first two podcast outputs felt disorganized — going off on tangents or not focusing on the most important parts of our posts. But the third try, with the right set of instructions, left us with a 7-minute conversation we were pretty happy with:

🎵 Click here to listen. 🎵

Benefits of Deep Dive

We can see why everyone was so impressed with the tool when it first started making headlines. Audio content can be hard for marketers, requiring a significant amount of hours and dollars. But this was fast, free, and totally intuitive to use.  

The content is engaging — it really does sound like a podcast! The output feels impressively human, introducing the key topics from our article inputs in a totally different way while maintaining a narrative structure. And because it’s based entirely on what we fed it, it’s less likely to hallucinate and go off track, as we know AI has a tendency to do. 

Drawbacks of Deep Dive

Outside of your original instructions, you have no control over the output at all: you can’t specify the length, you can’t edit the script, and you don’t really know what you’re going to get until after it’s been generated. As marketers, we’ve always got messaging as our north star, and not being able to guarantee consistency leaves us feeling a little wary. 

While the output is pretty lifelike, certain moments make it clear it’s coming from two disembodied voices. And these speakers, presumed experts on the topic at hand, are never identified by name. As consumers, we’re still getting used to what an AI-generated world could look like, and hearing one of the AI hosts refer to the other as “an expert” gave uncanny valley more than respected authority. 

We’re also keeping in mind that NotebookLM is aimed at students and researchers, not marketers. As a result, the functionality of the Deep Dive tool is not currently advanced enough to create high-quality, reliable content outputs. Google has acknowledged this, and is currently working on a business version of the tool. But until that rolls out, you can’t expect a workable piece of marketing content on the first try. Especially for businesses in regulated or specialized industries, even one brief, inaccurate detail could reflect poorly on your brand. 

Our AI-Generated World

In today’s impossibly fast-moving marketing environment, you have to stay on top of the latest tools or risk getting left behind. But not every trend will be applicable for your business. If quick, easy, and free audio content is something you’re looking for right now, NotebookLM might be a good choice. Some brands have already jumped on the AI audio train: Salesforce consulting partner 10K created a podcast companion for its annual Salesforce Talent Ecosystem Report, distilling the 20-page report into a six minute conversation. 

But if strict adherence to messaging guidelines, authentic content, and creative control is important to you, NotebookLM probably won’t fit your needs at this point. Instead, you could automate the audio portion of the podcast process with AI voice generators like ElevenLabs or use a text generator like ChatGPT to draft a script that humans can then record. 

Following the success of NotebookLM, we’re expecting to see more tools like it pop up and will be sure to report back when we find something worth trying.

Stay on top of the latest AI trends with Prompted, our biweekly blog series bringing you relevant updates and insights on how to strategically add AI-driven solutions into your marketing toolkit.