Latest / The CMO Podcast with Fexingo: Marketing Leadership, Budgets, and Executive Strategy / How CMOs Are Using Generative AI for Podcast Ad Creative
Transcript
- Lucas: Luna, I want to talk about something that's quietly reshaping how CMOs think about podcast advertising — and it's not about the cookie, it's not about targeting. It's about the creative itself. Luna: Okay, I'm intrigued. What's the shift? Lucas: Generative AI used to write and produce podcast ad scripts — at scale. I'm talking about a single campaign generating two hundred distinct ad variations, each tailored to a different audience segment, each with its own tone, hook, and call to action. A CPG brand I've been tracking did exactly that and saw a 34% lift in conversion versus their standard one-size-fits-all creative. Luna: Thirty-four percent — that's not marginal. But isn't there a tension there? Podcast audiences are loyal because they trust the host's voice. If you inject an ai generated script, doesn't that risk feeling fake? Lucas: That's the central tension, and we're going to get into it. But the technology is moving fast enough that some CMOs are already experimenting — and the results are forcing the rest to pay attention. Luna: Before we go deeper — if today's episode gives you a framework you actually use, the reason we can keep these conversations ad-free is listener support. You can buy us a coffee at buy me a coffee dot com slash fexingo. That's it — no upsell, just a way to keep the show independent. Lucas: Exactly. And we mean it — if you get value, that link is there. Now, back to the creative side. Let's talk about how the AI actually works. Lucas: Most CMOs are using fine-tuned large language models — think GPT-4 variants or Anthropic's Claude — trained on their brand voice guidelines, past ad scripts, and performance data. They feed in audience attributes like podcast genre, listener demographics, and the host's tone, and the model spits out a script in seconds. Luna: So the AI isn't just writing generic copy — it's trying to mimic the host-read style that makes podcast ads effective. Lucas: Right. And then the next step is synthetic voice generation. Companies like ElevenLabs and Respeecher can clone a voice from a few minutes of audio — with the rights holder's permission, hopefully — so you can have the AI script read by a version of the host's voice, even if the host never actually recorded it. Luna: That's where the trust issue really hits. Listeners are savvy. If they suspect the host didn't write or speak that ad, the credibility of the entire show can erode. Lucas: That's exactly the risk. And the CMOs who are doing this well are very careful about disclosure. The best practice I've seen is a clear audio watermark — something like 'This ad was created with AI assistance' — at the start of the spot. It's not hidden. And the brand is using AI to test variations, not to deceive. Luna: Okay, so let's get specific. Who's actually doing this today? Lucas: I talked to the CMO of a mid-market DTC brand — they sell premium meal kits — who started using a tool called AdCreative.ai for podcast scripts six months ago. They run about fifty podcast campaigns a quarter, each targeting different listener personas: busy parents, fitness enthusiasts, budget-conscious millennials. Luna: And before AI, they were writing each script by hand? Lucas: Exactly. One creative director, maybe two, writing each script from scratch. They could produce maybe ten to fifteen variations per campaign. Now with AI, they generate a hundred distinct scripts, test twenty in a pilot, and then scale the top performers. Their cost per acquisition dropped by about 18%. Luna: Eighteen percent CPA improvement — that's real money. But what about the host relationship? Did they tell the podcast hosts they were using AI? Lucas: They did. And that's critical. The hosts were given the option to review and tweak the AI scripts before recording. Most chose to do a quick edit for their own voice — adding a personal anecdote or a local reference. So it became a hybrid: AI draft, human polish. The hosts felt ownership, and the ads still sounded authentic. Luna: That feels like a smart middle ground. But I'm wondering about the Fortune 500 side. Is anyone doing this at scale? Lucas: Yes. One large retailer — think Walmart-level but with a bigger e-commerce presence — has been running a pilot with a synthetic voice platform. They're using it for programmatic podcast ads, where the ad is dynamically inserted. The AI generates a script, clones the voice of their in-house brand spokesperson, and then that voice reads the ad across dozens of shows. Luna: So they're not using the show's host voice — they're using a brand voice that listeners already associate with the company. That might actually preserve trust. Lucas: That's the theory. And their early data shows that listener recall for the ai generated brand voice ads is within 5% of human-recorded versions. So the synthetic voice quality is good enough that most listeners don't notice a difference. Luna: But what about the creative itself? Is the AI writing actually good? Or does it sound like generic marketing speak? Lucas: That's the rub. The AI can generate grammatically correct, on-brand copy, but it often lacks the specific detail that makes a podcast ad memorable — like a host mentioning their own experience with the product. The best performing AI scripts in that CPG campaign were the ones that included a simulated personal story. The model was fine-tuned on examples of hosts sharing real anecdotes. Luna: So it's not just generating copy — it's generating fake authenticity. That's a different ethical layer. Lucas: It is. And that's why I think the CMOs who win with this will be the ones who use AI as a creative partner, not a replacement. They'll have the AI generate a dozen hooks and a framework, but then a human editor — ideally the host — adds the actual experience. Luna: What about measurement? How do you know if an ai generated ad is actually more effective than a human-written one, beyond the CPA? Lucas: Great question. Some brands are running controlled A/B tests within the same podcast — same audience, same show, same slot — comparing an AI script read by the host versus a human-written script read by the host. The early results are mixed. In one test, the AI script outperformed on click-through rate because it had a sharper call to action. In another, the human script had higher brand recall because it included a specific anecdote. Luna: So the answer isn't 'AI always wins' — it's about using AI for the parts it's good at and keeping humans for the parts that need real connection. Lucas: Exactly. And that's the framework I want to leave listeners with. If you're a CMO thinking about this, here's a practical way to slice it. Lucas: Use AI for three things: first, generating a high volume of script variations for testing. Second, personalizing ad copy for different audience segments at scale. Third, creating synthetic voice ads for programmatic placements where the host isn't involved anyway. Luna: And keep humans for the stuff that builds trust — the host-read ad where the host talks about their own experience, the brand voice that listeners already know, the final quality check before anything goes live. Lucas: Right. And one more thing: disclosure. If you use synthetic voice or ai generated scripts, tell the listener upfront. Not just because it's ethical — because if they find out later, the trust damage is worse than any CPA improvement. Luna: So where do you see this going in the next year? Are we heading toward a world where every podcast ad is ai generated? Lucas: I don't think so. I think the premium will be on authenticity. The shows that have the strongest host-audience trust will continue to command higher CPMs for human-read ads. AI will fill the middle and lower tiers — the programmatic inventory, the niche shows where a host-read ad doesn't justify the production cost. Luna: So the market bifurcates: premium human ads at a premium price, and ai assisted ads for scale. Lucas: That's my bet. And the CMOs who figure out the right mix — where to use AI and where to invest in human connection — will have a real advantage. It's not about replacing the host-read ad. It's about making every ad dollar work harder. Luna: That's a good place to leave it. Thanks, Lucas. Lucas: Thanks, Luna. If you want to dig into the numbers from that CPG campaign, we'll link the case study in the show notes.