AI podcasts have quickly gone from curiosity to flood.
Over one recent nine-day stretch, nearly 40% of new podcasts were likely AI-generated, according to Podcast Index, as reported by Bloomberg. One company alone, Inception Point AI, now claims to have more than 10,000 active shows, with hundreds more appearing almost overnight.
This is publishing at industrial scale. The real question is not whether AI can make podcasts. It clearly can. The question is whether people will consume them.
Insights From a NYU Focus Group
Last fall, I asked my NYU Business of Podcasting students to evaluate AI-generated podcasts. The great media researcher Tom Webster joined that class, turning it into a real-world focus group. This spring semester, I ran another test. Students selected podcasts on topics they already cared about, which brought some knowledge and standards to the experience.
I asked them to listen and rate the podcasts on a 1-to-5 scale.
Across the board, the shows averaged a 2.3 out of 5 rating. That score does not mean the podcasts failed completely. Students recognized strengths. They reported the shows were often structured, efficient, and a handful were useful. Some delivered information clearly. In narrow utility settings — a weather update, a quick explainer, a product summary, AI audio can absolutely serve a purpose.
But information is not the same thing as experience. Students repeatedly described the shows as robotic, monotone, flat, or like “audio ChatGPT.” Most did not feel a connection. Many, but not all, did not want to come back to another episode.
One student listened to an AI-generated surf report and essentially said it delivered what was needed, but it was not enjoyable.That distinction likely matters in the discussion about AI podcasts. Utility audio solves a task. Great podcasts create attachment. A podcast can inform without earning loyalty, deliver facts without becoming memorable, and function efficiently without giving listeners a reason to return.
AI Podcasts Aren’t One Thing
There is plenty of industry talk about whether AI podcasts should somehow be banned. Not so easy, and not so wise.
Alberto Betella of RSS.com was on The PodNews Weekly Review and offered useful ways to think about AI and podcasts: not all AI podcasts are the same and treating them as one category muddies the issue. There are AI-assisted podcasts where humans use tools to improve production. No problem. There are AI-curated utility podcasts delivering structured information. Seems fine.
Then there is spam designed to manipulate systems. This is the podcast version of fake handbags with familiar-looking names, cover art and feeds created to confuse, capture a click, or game search returns. That is simply deceptive and horrible.
Then there is what many are calling podslop: fully automated shows produced at scale with little human judgment and even less accountability.
Betella’s useful point is that AI slop is not binary. It’s a spectrum. A weak AI podcast about golf may be annoying. A weak AI podcast offering health advice or financial guidance can become something else entirely. The point is AI itself is not the concern, it’s the potential impact.
Podcasting Runs on Trust
Podcasting works because it feels human.
It is a voice you come back to, a perspective you recognize, and a sense that someone has done the work and is speaking with intention. This is where many AI podcasts break down. It can mimic tone, but rarely carries conviction, personality, or a natural human rhythm – a giggle, or an “oh my.” The delivery often feels flat and emotionally empty.
Scale Over Substance
Another important factor as we think about all of this.
Podcasting has always been powered by the longtail of niche topics. Many of the AI shows are built around even more narrow niches and micro-topics that may never justify a human-hosted show. That might work.
But niche shows are hard to scale so it makes sense that much of what is being produced today out of AI farms appears to be about tonnage.
Programmatic advertising makes that model possible. A small audience and a low CPM may not matter much on one show. Multiply that by thousands of feeds and the math begins to look different. It may be nickels and dimes for now, but nickels and dimes add up quickly when the cost of production is close to zero.
That is where brand safety enters the picture. Advertisers should know where their messages are running. If a brand appears inside a lightly reviewed AI health show, a synthetic finance podcast, or a podcast with no visible human accountability, the issue is not just reach.
The Audio ChatGPT Problem
As I mentioned, several students independently arrived at the conclusion that these shows sounded like audio ChatGPT.
That’s not simply an insult; it’s a useful diagnosis.
Large language models are very good at organizing and synthesizing. But successful podcasts depend on pacing, tension, timing, surprise, chemistry, humor, taste, originality, and point of view.
One student noted that an AI narrator described a brutal crime with the same emotional range someone might use to describe breakfast. Another student said: ‘The ad had more enthusiasm than the storyteller.’
The technology can produce speech. But speech is not performance. And podcasts, especially the durable ones, are about performance.
Gen Z and the Authenticity Gap
Gen Z understands AI. They use it. They expect it. It may surprise some of you, but that does not mean they want everything to feel machine-made.
In a world that is increasingly automated, authenticity is becoming more valuable by the day. Listeners and viewers respond to intent, perspective, and the sense that someone made choices about what to say and how to say it. They’re also becoming better at detecting when something is synthetic, or hollow.
That’s why my students’ response is so important. This is not a group that is frightened by technology or nostalgic for the old way of doing things. If anything, they are the most AI-native audience we have. Yet they still recognized the gap between useful summaries and a satisfying listening experience.
Trust Is the Real Battleground
What surprised me most was not simply that students criticized robotic voices. It was how quickly they moved to questions of authenticity and credibility.
One student listened to a mystery podcast, searched for the real case afterward, and concluded the story appeared to be fabricated. Her reaction was not just that the show was weak. It was that she did not understand why anyone would want to listen to a made-up story presented in that way. It’s a fair point.
Once trust erodes, rebuilding it is hard.
Start with Disclosure
The industry response to AI-generated podcasts is still early and uneven.
There’s a temptation to overcorrect by removing AI-generated content altogether. Some platforms and aggregation sites may decide to exclude anything created by a machine. I’m not a fan of that. We should not make decisions for the audience. Let listeners decide what is valuable and what is not. Automatically wiping it out ignores both AI’s potential value and its inevitability.
A more practical place to begin is disclosure. Identifying when a portion of a podcast is AI-generated creates transparency for listeners, flexibility for advertisers, and a basic framework for platforms.
Podcasting is not alone in confronting this issue. Spotify is reportedly exploring verification and disclosure systems as synthetic audio spreads across its platform. YouTube already requires creators to disclose certain types of AI-generated or altered video content. Across media, the industry is slowly converging around the simple idea that audiences should know when synthetic content is shaping what they hear and see.
Regulation is moving in this direction as well. The European Union’s AI Act will require more transparency around AI-generated content.
Podcasting has an opportunity to define standards before they are imposed externally.
Attention is Still Scarce
We have already had a tsunami of content, and AI intensifies it. More shows. More feeds. More synthetic voices. More competition for the same finite amount of time.
AI lowers the cost of making content, but it does nothing by itself to increase the value of that content to a listener or viewer. If anything, it increases the noise.
Not all AI podcasts will fail.
But audience time is still finite.
The future will not belong to whoever can generate the most content. It will belong to whoever creates something people repeatedly choose and genuinely want.
The mission hasn’t changed: make something worth paying attention to. That’s what keeps people coming back.