There is a strange little corner of the podcast world called Fexingo.

At first glance, it looks like a podcast network, but delver deeper and you soon realize it’s nothing like most podcast networks.

Its own website describes it as “a universe of podcasts” with 755 shows across language, history, horror, business, finance, marketing, technology, careers and economics. There are 300 history podcasts, 100 horror podcasts, 55 language podcasts, and so on, and so forth.

Most are hosted by two recurring voices, Lucas and Luna, who politely guide listeners through short, educational conversations that feel as natural as a deep tissue massage from C3-P0.

That’s because this beast of a production house is totally AI. A machine for making shows, not one big show or a single carefully developed title passion project, Fexingo is building a universe of content.

Fexingo does not appear to plainly say, “Hello, these are AI-generated podcasts” when you stumble across them on your app or player. Yes they have AI generated art, but that seems common for organic made show these days anyway. So listeners are going to stumble onto these shows unknowingly, expecting a real person and having the experience of either being fooled or experiencing a gradual realization.

So lets experience it from the listeners perspective.

With an upcoming vacation to the German and Austrian Alps, I chose to listen to The History Of Austria: Empire, Collapse, and Reinvention- Fexingo History.

Lucas and Luna publish two episodes a day, each about 5-6 minutes long, and are on episode 132 despite launching in late April. I swallowed my pride, scrolled to episode 1 and hit play on this adventure through time.

The music kicked in, Lucas gave an well paced intro and then Luna chimed in with some “banter.” Let’s just say that despite all this time working together, our hosts are yet to develop any sort of chemistry.

History comes alive when those who study it and share it, ooze with the passion that drives their work. I love when a history show is powered by a voice that has obsessed over a hyper-specific period or moment that infects the listener. There is no risk of that with this show.

It was informative, and seemed accurate, but I knew deep down it hadn’t been “researched.” In the back of my mind, how much was real history and how much was AI hallucination? Could I really trust this as a reputable source? I know it hasn’t been checked based on the rapid release schedule.

Weirdly, Luna claimed she’s “walked the walls” of an Austrian town and commented on how beautiful they were. Not only am I sure that she’s lying, but that just unnerved me. Just be honest about being AI and give me the information, don’t make stuff up in a bizarre attempt to emotionally engage me based on manufactured experiences. Cold shivers.

This always hits home with real people, when global history professor Peter Frankopan describes a memory from visiting an ancient city, I feel like I’m there. When Luna does it, I start to doubt reality and wonder if that site ever existed.

To summarize, this is “technically” a historical podcast. “Technically” tomatoes and cucumbers are fruits, but I’m yet to find a yogurt that flavour. It is capable of making content that ticks the basic boxes, but failed miserably in generating the spark that makes a show connect emotionally. The thing that actually compels us to listen.

What’s the good and the bad?

This is not just about whether AI shows are good or bad. Some will be useful, some will be harmless, some will be genuinely impressive. There is nothing inherently noble about a bad human podcast, and nothing inherently worthless about a synthetic one.

I actually had fun making one myself about two years ago. The voice is AI, but I did write the script and edit it. It was a silly and pointless experiment that went as far as a single two-minute episode, but experimenting is important for so many reasons.

Also, and more importantly, from an accessibility point of view, AI generated voices can unlock so much new content for the visually impaired or even for dyslexics like myself. There are benefits.

For years, the promise of podcasting was that anyone could make one. All you needed was an idea, a microphone, and enough stubbornness to keep going when only 34 people listened to episode three.

But when one person can make 755 podcasts, you get a discovery crisis.

When one person is producing hundreds of shows, with recurring synthetic-sounding hosts, daily updates, enormous category coverage and an endlessly expanding content library, it is not unreasonable to say this looks like an AI-assisted or automation-heavy podcast operation.

Podcasting has always had a messy discovery problem. It is still far too hard for a great independent show to be found by the right listener. Now imagine search results, category charts and platform recommendations filling with thousands of calm, competent, keyword-friendly shows on every imaginable topic.

Unless we start seeing platforms give AI specific labels to shows like this, our only hope is to “be more human.”

The opportunity is to be organic, chaotic and more human than ever.

Make shows that feel authored with a centre of gravity anchored in the human experience. The listener needs to understand who is speaking, why they care, and why this conversation could not have been generated from a prompt and a content calendar.

For brands, that means no more beige thought leadership dressed up as a podcast. Your show needs a reason beyond “we want to be in the space.” It needs a host with curiosity, personality and editorial tension.

For independent creators, it means your weirdness is not a liability, it is the moat. Your obsession, your humour, your taste, your local references, your in-jokes, your community, your slightly-too-long story about the thing that happened in 2008. That is the good stuff. That is the signal that we are real.

For platforms, this is the harder bit. Podcast discovery cannot keep pretending all audio is equal just because it has an RSS feed and a description. If automated networks can generate hundreds or thousands of shows, then directories will need better ways of identifying originality, authorship, trust and engagement. Otherwise, search becomes a landfill we must dig through to find gold.

Fexingo may be the start of something much bigger. It may be one person’s ambitious experiment in building a giant library of useful audio. It may also be an early glimpse of what happens when both the cost of making a podcast, and the commitment to audience experience gets close to zero.

AI can make podcasts. But can it make listeners care?