Google’s new tool, NotebooKLM, creates uncannily realistic podcasts, sparking intrigue and concerns over AI’s role in content creation

AI has come a long way, but no one could have predicted it would make its mark by flooding the world with even more podcasts. Google’s latest tool, NotebookLM, has made this all too real. Like a scene from a dystopian sci-fi film, where robots dominate every corner of human life, AI now promises to take over one of humanity’s most beloved modern hobbies: podcasting.

For years, the internet has groaned under the weight of endless podcasts. Celebrities like Gyles Brandreth and Paul Giamatti have joined the fray, and even your least interesting friends have probably spent hours recording their own podcast episodes. But now, Google has upped the ante, introducing AI-generated podcasts that can be produced at the click of a button. The results are not just fast—they’re eerily realistic.

NotebookLM functions similarly to ChatGPT, but instead of producing text, it creates audio content. You can upload sources such as documents, websites, or YouTube videos, and the AI crafts a podcast, with two hosts—one male, one female—discussing the chosen topic. What sets these AI hosts apart is their uncanny resemblance to human speech patterns. They fumble with their words, they talk over one another, and they pause awkwardly as if thinking. It’s all so natural that, for a moment, you forget you’re listening to robots.

The world, of course, isn’t crying out for more podcasts. As the author dryly puts it, “The world needs more podcasts like it needs to be kicked by a horse.” But the rise of AI-generated podcasts poses fascinating questions about the future of content creation.

NotebookLM bills itself as a study tool, designed to help users condense and summarise information into digestible audio that can be consumed on the go. For students prepping for exams or anyone cramming large volumes of data, this could be a game changer. But the implications stretch far beyond education. The ability to generate podcasts on virtually any topic opens up endless possibilities. Curious to see what it could do, the author fed the AI tool interviews related to Rivals, a Disney+ show that has been making headlines for its throwback to 1980s culture.

The five-minute podcast that NotebookLM produced was eerily slick. The female host excitedly opened the discussion with, “All right, get ready, because we are diving into Rivals!” Her male co-host enthusiastically agreed, and together they discussed the show’s approach to 1980s sexism and racism, portraying Rivals as a groundbreaking, agenda-setting series. However, as the author points out, the show is far from the serious social commentary the AI made it out to be—it’s actually a campy, nudity-filled romp. The AI, though, had only pulled from interviews with the actors, who understandably focused more on social issues than their on-screen antics.

This experiment highlighted a key flaw: the podcast was confident, natural, and articulate, but ultimately based on selective information. It could have been more accurate had it included diverse sources—interviews, reviews, the source novel, and more. But then again, aren’t all podcasts guilty of being selective? Many are built on the same foundation: limited information is presented with absolute confidence.

The author’s next test of NotebookLM took a different approach. This time, they chose the well-worn genre of conspiracy theory podcasts. Though it could have been filled with the wildest theories from flat-Earth forums and anti-5G Facebook groups, the author chose to throw in a mix of Wikipedia and Reddit content. To their surprise, the AI-generated podcast was thoughtful and measured, offering a balanced 14-minute discussion about confirmation bias and the human need to explain the world. At one point, the male host cut short a list of conspiracy theories, quipping, “My brain would literally explode.” The tone was unexpected and oddly insightful.

While the podcasts generated by NotebookLM are fascinating, they also raise important questions. Will this AI tool make people too lazy to do their own research? Can the content it produces be trusted when it draws from limited or biased sources? And what happens to the millions of podcasters already saturating the market? As AI continues to blur the line between man and machine, these questions are becoming more urgent.

But there’s no denying the brilliance of NotebookLM in its current form. Whether it’s summarising a complex topic or producing a natural-sounding conversation, this tool is revolutionary. It’s another step in the AI evolution, offering a glimpse of how technology can transform even the most human of activities—such as two people chatting about nothing for 30 minutes—into something created by a machine. So, while AI might not be coming for us in the terrifying ways that The Terminator or The Matrix imagined, it is poised to take over our earbuds, one automated conversation at a time.