Steve Goldstein’s Amplifi Media works with media companies and podcasters in developing audio content strategies. Goldstein writes frequently at the Amplifi blog. Steve can be reached directly at 203-221-1400 or sjgoldstein-at-amplifimedia-dot-com.
If you’re in the audio business and haven’t yet explored Google’s new NotebookLM audio tool, you should. It’s an impressive, jaw-dropping tool that is exciting, weird and unsettling at the same time.
I tried it. I took an Amplifi Media Thought Letter post from earlier this year about subscription podcasts uploaded it and within minutes NotebookLM generated a rich six-minute conversation between two “people” about the topic and my post.
I played this audio for some of my colleagues and my Business of Podcasting class at NYU. The episode features two hosts expanding on my blog and their thoughts on the podcast companies mentioned. Their conversation includes natural banter like “yeah,” “oh wow,” and “like,” with realistic human intonation. They interrupt each other and ask questions. Everyone was shocked to learn that the entire episode was AI-generated. All I did was upload my post. NotebookLM did the rest.
How It Works
Marketed initially as a study tool, NotebookLM now offers an “audio overview” feature that creates AI-generated podcast episodes. Imagine turning any article, video, PowerPoint deck, research chart, or even a loose collection of ideas into a compelling two-person conversation. That’s what NotebookLM can do.
In minutes, it takes about two clicks to turn any content into a rich podcast. Instead of a monologue or a flat narrative, it structures the episode as if two individuals bounce ideas off each other, offering differing perspectives and elaborating on key points. The AI-created dialogue took my thoughts and turned them into a lively discussion with a relatable tone. Google calls these “deep dives,” and they can cover any topic. The unnamed (ahem, they are not real) male/female voices interject naturally, sounding uncannily human.
Companies are Pursuing AI Voice Content
Some businesses are embracing AI audio creation today. This development is the natural progression of voice. Seven years ago, I invested in SpokenLayer. This startup rode the Alexa wave, turning print and web articles into audio for top newspapers and magazines, including Time, The Economist, and others. The early AI voices were flat and monotone, which made the listening experience difficult, so we opted for professional narrators. Other companies like Audm followed. It was purchased four years ago by the New York Times and is now part of “New York Times Audio” using synthetic voices.
A few years later, here we are with AI creating better narration and lifelike conversation. NotebookLM is not the only audio AI. Many companies are pursuing it each with a different angle, and they are all impressive, but the Google product moves the bar higher.
-
Spotify has also been experimenting with AI voice translation.
-
Amazon Ads just launched new AI tools for advertisers: an “AI Creative Studio and Audio Generator.”
-
Amazon’s Audible is letting narrators clone themselves. There has been some backlash from some book narrators and consumers.
-
Descript offers voice cloning or stock AI voices.
-
Podcastle is an AI-powered suite of services that recently raised $13.5M in a series A funding round. Among its many products, Podcastle offers Revoice, an AI voice cloning tool.
Should Podcasters Worry About Google’s NotebookLM?
NotebookLM is making noise, and people are trying it. The Wall Street Journal wrote, “There’s a new hit podcast that will blow your mind.”
Yes, it has a certain cool about it, but is this a threat to podcasters? Is the content that good? Is this another sea change in the podcast space, or is this the next flash-in-the-pan like the pandemic hit Clubhouse?
Let’s put NotebookLM on the side for a moment. AI audio must be taken seriously. The implications of AI podcasts are right up there with bogus videos, pictures, impersonations, and other questionable content. As the tools improve, distinguishing between real and fake becomes progressively harder. This challenge extends far beyond podcasting, with companies wresting with the implications for political campaigns, business, personal likeness, and the spread of misinformation.
Wenbin Fang, creator and CEO of the podcast search engine Listen Notes, published a long essay titled “NotebookLM: A Threat to the Podcasting World” in defense of original content and announced they will not track AI podcasts from Google because it clutters “the podcasting space with low-quality, automated content that is not intended for genuine human listeners.”
I’m not completely with you Wenbin. For a search engine, this is indeed tricky, but these are not fake Taylor Swift videos. To be sure, there are plenty of ethical issues with AI, but genuine human listeners should decide whether this content is right for them. Indeed they may not.
Listen Notes even built a tracking tool which, as of this writing, has just over 1,200 NotebookLM “podcasts.”
Like all AI, NotebookLM can hallucinate or even nuttier. Listen to this snippet when someone sends the system into an existential spiral when the “hosts” realize they are not human.
Very meta.
How These “Podcasts” Will Likely Be Used
NotebookLM was presented as an educational aid. Thinking of my role as a professor at NYU, NotebookLM might be a great tool for educators looking to break down complex ideas into digestible bits to make learning more engaging. I intend to try it.
Translating that into the business world, companies might take those ignored policy memos and make them into more entertaining content.
They might make for great employee training, customer support, or HR summaries.
“The magic of the tool is that people get to listen to something that they ordinarily would not be able to just find on YouTube or an existing podcast” said Raiza Martin, who leads the NotebookLM team at Google Labs.
It likely will never capture the rich sound and rhythm of human voices riffing, trash-talking, debating, and connecting with listeners
What Should Podcasters Do?
Should podcasters be quaking in their boots? Probably not. While the AI capability is scary and impressive it likely will never capture the rich sound and rhythm of human voices riffing, trash-talking, debating, and connecting with listeners. AI can’t replicate the authentic spontaneity, personality, and laughter that makes many podcasts engaging and keeps audiences returning for more. That kind of magic is hard to automate.
Even at its best, AI content often feels just a bit off.
The human factor endures and remains essential in all media. But let’s be clear: if you listened to the podcast that started this post, it’s obvious that NotebookLM has raised the bar. It is remarkable technology. If you’re creating average, mediocre, or dull content this is another blow.
Human Connection
NotebookLM is undeniably fascinating—exciting, unsettling, and a little eerie. While tools like this can amplify or streamline production, the essence of lasting content comes from the human touch. It’s in the stories crafted by real voices, the authenticity that resonates, and in the connection that only a person can make—like the irreplaceable impact of eye contact. True engagement isn’t built on automation but on the enduring bond between creator and audience. Whether this is another fleeting tech fad like Google Glass or Segue or a transformative tool like ChatGPT, only time will tell. But human connection will always be the anchor that holds.