AI Podcasts: Cheap, Synthetic, and Everywhere

Lachlan Reed

[excited] Welcome to the show! I'm Lachlan Reed with Simon Carver, and mate, I need to start with the number that made me nearly drop a spanner in the shed: an AI-made podcast episode can now cost around ONE dollar to produce. One buck. That's not "cheap for media" cheap -- that's vending-machine cheap.

Simon Carver

[questioning tone] One dollar is the part that gets me. Not because it's futuristic, but because it changes the behavior. If an episode costs a studio thousands, you deliberate. If it costs a dollar... you don't deliberate, you spray and pray. You make a hundred. A thousand. You see what sticks.

Lachlan Reed

[amazed] Exactly. And that's not hypothetical anymore. Some AI studios are already pumping out HUNDREDS OF THOUSANDS of episodes. That's the bit that shifts the ground under your feet. We used to think a podcast was, you know, two people with mics, a producer, maybe a dodgy spare room with egg cartons on the wall. Now it's scripts, cloned voices, auto-editing, thumbnail, title, publish -- done before you've made a cuppa.

Simon Carver

[reflective] "Hundreds of thousands" is not a cottage industry. That's not even a busy network. That's a content weather system. And I think the strange part is that the word podcast still sounds intimate, handmade, almost confessional. But at that scale, it starts to feel less like a conversation and more like a factory line pretending to smile at you.

Lachlan Reed

[laughs softly] A factory line with a soothing voice and perfect mic technique. And one example people keep pointing to is Inception Point AI. Their whole thing shows how speed and scale have gone a bit bananas. Synthetic hosts, rapid production, big volume. It's not just, "AI helped tidy the audio." It's, "AI can BE the host." That's a different beast entirely.

Simon Carver

[stunned] Wait -- "synthetic hosts" is the phrase I can't let go of. Because host used to mean person. A host was someone whose pauses meant something, whose weird laugh you recognized, whose bad day you could hear even when they tried to hide it. If the host is synthetic, what exactly is being hosted? My attention? My trust?

Lachlan Reed

[curious] Yeah, and that's where I get a bit tangled, because part of me thinks, well, if the episode is useful, maybe listeners don't care. If you need a five-minute market update or language lesson, maybe a synthetic voice is fine. Like GPS. I don't need my sat-nav to have a childhood. I just need it not to send me into a lake.

Simon Carver

[responds quickly] I think that's true for directions. I'm not sure it's true for companionship. A lot of podcasts live in that odd middle space -- information, yes, but also company on a walk, company in traffic, company while you're folding tiny socks at midnight. If I think I'm building a relationship with a person and it turns out I'm bonding with a system... that feels different. Not evil, necessarily. But different in a way that matters.

Lachlan Reed

And listeners often can't tell who's human anymore, or not straight away. That's the trust problem in a nutshell. The old bond was pretty simple: you hear my voice, I say "this is me," and unless I'm being a drongo, you assume there's an actual bloke or person on the other end. Now that assumption's gone a bit pear-shaped.

Simon Carver

[skeptical] Though let me push you a little. Haven't we always had performance? Radio voices, stage names, heavily edited chats, producers feeding questions in the ear. Podcasts were never pure, untouched humanity in a mason jar. So is the problem really fakery... or is it undisclosed fakery?

Lachlan Reed

[pauses] Ah, that's sharper. Because you're right -- media's always had polish. But there's a difference between editing my ramble and inventing the rambling person. If a show says, "These hosts are AI-generated," okay, cards on the table. Weird, maybe useful, maybe not for me. But if it PRESENTS as human and trades on warmth, vulnerability, little stumbles, all that good stuff... then we're in sticky territory. Even a kangaroo could trip over that one.

Simon Carver

[softly] And once listeners start wondering, "Are you real?" they don't just ask it about the suspicious shows. They ask it about everyone. That's the part I find sad. The flood doesn't only create more content. It erodes the background assumptions that made intimate media feel intimate in the first place.

Simon Carver

So then the obvious fear is replacement. Human podcasters versus the machine, last one standing, cue dramatic music. But when I talk to actual creators, the first use cases are usually much less cinematic. It's transcripts. It's cleaning up audio. It's translating episodes. It's drafting show notes. It's chopping one long interview into six decent clips you can post without losing your weekend.

Lachlan Reed

[warmly] Yep. The back-end stuff first. That's where AI feels less like a robot stealing your sandwich and more like an intern who never sleeps. Editing out filler words, leveling sound, making captions, pulling timestamps. For heaps of creators, especially small ones, that's the difference between publishing consistently and just... not. Because the glamorous part is talking into the mic. The non-glamorous part is the six fiddly jobs afterward.

Simon Carver

The "six fiddly jobs" is so real. And translation is a big one. If AI can help a show jump from one language to another, that's not trivial. That's access. Same with transcripts for search and accessibility. So I don't want to flatten this into "AI bad, humans good." A lot of the practical uses are honestly pretty humane. They give people time back.

Lachlan Reed

[matter-of-fact] Time back is the currency, hey. If a creator saves three or four hours a week on admin sludge, that's three or four hours they can spend reporting, writing, interviewing, or just thinking. And thinking matters. Taste matters. Choosing the right question matters. AI can repurpose a clip, sure, but it can't tell you why THAT clip is the one that hits a nerve -- not in the same lived way.

Simon Carver

[curious] But here's where I wobble a bit. Generative AI use is rising fast across the creator economy -- that much is clear. Yet people still say they trust human-made content more, especially when the stakes are high. Health advice. Politics. Money. Grief. Anything where the words are not just entertaining but consequential. So what happens when the tools get better than our instincts? When the machine-made thing sounds more polished than the human-made thing?

Lachlan Reed

Then honesty becomes the whole ball game. Not polish. Not efficiency. Honesty. If I use AI to clean up audio, fine, who cares. If I use it to generate show notes from my own episode, also fine. If I use it to dub my voice into another language and I SAY that, sweet as. But if I let the machine write feelings I didn't feel, or fake an interview, or create a synthetic version of me to simulate intimacy... nah. That's where the ute tips over.

Simon Carver

[reflective] "Write feelings I didn't feel" -- that's the line I'm going to remember. Because the upside here is real. Friction disappears. More people can make things. Small teams can sound bigger. Independent creators can compete with studios on workflow. That's exciting. It could widen the circle of who gets to speak.

Lachlan Reed

And maybe even help niche creators survive. A history show with two thousand loyal listeners, a local footy show, a disability advocate translating episodes, a teacher turning lessons into audio without hiring a whole crew -- that's genuinely good stuff. The machine can handle the grunt work. The human keeps the judgment, the scars, the weird side stories, the point of view.

Simon Carver

[skeptical] Unless the convenience slowly creeps, right? First it's transcripts. Then summaries. Then suggested questions. Then a draft intro. Then an AI co-host to fill space. None of those steps alone feels outrageous. But stack enough of them and the creator might still be present while somehow becoming less... authored.

Lachlan Reed

[hesitates] Yeah... that's the quiet risk. Not replacement in one dramatic whack, but erosion by convenience. You save time, beautiful. Then one day you realize the machine isn't just helping you say the thing -- it's steering what gets said because it's faster to accept the suggestion than wrestle your own thought into shape.

Simon Carver

And maybe that's the question hanging over all this. If AI gives creators more room to focus on voice, taste, and original ideas, brilliant. But if voice becomes style, and taste becomes optimization, and "original" just means lightly edited machine output... how would a listener know when the human ends?