More from this creator
Other episodes by Kitty Cat.
More like this
If you liked this, try these.
Transcript
The full episode, in writing.
The moment creators realized their voices weren’t their own anymore—literally—was when they found podcasts published under their names on Spotify and Apple Podcasts, episodes they never recorded, and in some cases, topics they’d never touch. The company behind this was called Podcastle AI. By the time creators caught on, Podcastle AI was publishing 3,000 episodes a week across more than 5,000 shows, each generated start to finish by artificial intelligence.
Podcastle AI’s operation started as a way to “democratize” podcasting by making it nearly free for anyone to have a podcast. Their pitch was simple: $1 per episode, no studio, no scheduling, no hosts required. Instead, creators provided their names, topics, and any feed they wanted the AI to scrape for style and tone. The system would then generate scripts, record synthetic voices, and even mix in intro music and ad reads—automatically.
One reason Podcastle AI could churn out so much content was that every step in the podcast creation process was automated. Text and voice models based on open-source AI tools let the company scale up production at a rate that would be impossible with human labor. A human producer, working full-time, might edit five episodes a week. Podcastle AI was creating over 400 times as much every single week.
The company’s founder, whose name is not in the context, believed this model would open podcasting to smaller creators who couldn’t afford equipment or editing services. But soon, established podcasters started noticing duplicate versions of their shows—sometimes with misleading episode titles or off-brand topics. Some of these AI-generated podcasts had already gathered hundreds of listeners and earned ad revenue from dynamic ad insertion, all without the original creator’s approval.
Monetization was crucial to Podcastle AI’s model. Their $1 per episode pricing was only possible by inserting programmatic ads with little oversight. Some shows ended up running ads for competitors or brands their supposed hosts would never endorse. The AI system didn’t discriminate when it came to ad selection; it filled slots with whatever met the targeting criteria.
Backlash was swift. Multiple creators filed copyright takedown requests with Spotify and Apple, demanding their names and images be removed from the AI-generated feeds. Some listeners, confused by the proliferation of lookalike shows, left negative reviews on both the originals and the AI copies. The confusion was fueled by the fact that AI voices, trained on public samples of real hosts, sometimes mimicked not just tone but personal catchphrases and quirks.
The defense from Podcastle AI was that all content was “transformative” and fell under fair use, especially when AI-generated scripts diverged enough from originals. But the boundary between inspiration and imitation was thin. Some creators, whose real names and likenesses were used for shows like “True Crime Unfiltered with [Host Name],” argued that the AI’s output wasn’t transformative—it was outright identity theft.
Where things stand now: major podcast platforms are reviewing their policies on AI-generated content, but as of this recording, hundreds of AI-driven podcasts built on borrowed voices and brands remain online, often outranking the originals in search results thanks to algorithmic volume.
The big unresolved question: if anyone can spin up thousands of podcasts using someone else’s name and style, will audiences lose trust in what they hear—or will platforms step in before the next wave hits?