Yesterday I guest lectured at Steve Goldstein’s wonderful Business of Podcasting class at NYU, and one of the assignments he gave his students for us to discuss was to listen to an AI-generated podcast and give their feedback. I’ll have more to say about that class soon (Steve and I are collaborating on a post), but one thing that set my brain on fire after that class was a question I’ve been grappling with for weeks now: given that so much of the trust we have in podcast hosts (and in the ads they read) is based on parasocial relationships, can we have a parasocial relationship with an AI?
I spent hours digging for answers to this question, and luckily I found some very recent work on the topic in a journal called Computers and Human Behavior, from a fascinating paper from the School of Economics & Management at Changsha University. It’s a long-ish paper called An Assistant or A Friend? The role of parasocial relationship of human-computer interaction. The researchers wanted to understand how humans develop parasocial relationships with AI – you know, those one-sided emotional connections we form with media personalities. Think of how you feel about your favorite podcast host. You probably feel like you know them, even though they have no idea you exist. It’s the same psychological mechanism that makes people send wedding invitations to Oprah.
Here’s where it gets interesting: the study found that we don’t just form one type of relationship with AI. We form two distinct types – “assistant” relationships and “friend” relationships. And which one we develop dramatically affects how we perceive that AI’s competence and warmth.
When people saw AI as an assistant, they perceived it as highly competent but less warm. When they saw it as a friend, the warmth scores shot up while competence took a back seat. It’s like the difference between how you’d react to Siri versus how you’d react to your buddy Chad.
The study also dove into something that makes a lot of people uncomfortable: gender traits in AI. Male-presenting AI assistants were perceived as more competent. Female-presenting AI friends were seen as warmer. Yes, these are stereotypes, and yes, the makers of smart speakers and voice assistants already knew this.
The research suggests that an AI version of a host’s voice reading ads might actually be more effective than we think – but only if it maintains the right parasocial relationship type. A tech podcast with an “assistant” vibe? That AI voice better nail the competence factor. A true crime show where hosts have built that intimate “friend” relationship with listeners? Warmth is non-negotiable.
Trust Is The Real Currency
Here’s the kicker that should have every podcast advertiser paying attention: both perceived competence and perceived warmth positively influenced consumer attitudes. But they work through different mechanisms. It’s not just about whether listeners trust the content – it’s about what kind of trust they’re experiencing.
Think about your favorite podcast ads (yes, I know you have them). The ones that work aren’t just competently delivered or warmly presented. They match the parasocial relationship you’ve already established with that show. This is where AI could actually enhance rather than erode trust in podcast advertising. Imagine programmatic ads that aren’t just dynamically inserted but dynamically performed – matching not just the host’s voice but the specific parasocial relationship type of that show. Your favorite conversational podcast maintains its warm, friend-like ad reads even across thousands of different geographic campaigns. That authoritative business podcast keeps its competent, assistant-like tone whether it’s selling software or socks.
Let’s address the elephant in the room: most podcasts don’t make money. The long tail of podcasting is littered with brilliant shows that can’t monetize because they can’t scale host-read ads. But what if every podcast in the tail could have host-read-style ads that actually match their unique parasocial relationship with their audience? Not generic programmatic spots, but AI-performed ads that maintain the show’s specific vibe?
The middle class of podcasting has been getting hammered for years – too small for major brand deals, too big to ignore monetization entirely. AI hosts reading ads might be their lifeline, but only if we get the parasocial dynamics right.
What This Actually Means For Podcasters
If you’re a podcaster reading this and feeling that familiar knot in your stomach about AI replacing you, take a breath. This research actually validates something you already know: the relationship you’ve built with your audience is the whole ballgame. AI can’t create that relationship from scratch. What it might be able to do is help you scale that relationship in ways that were previously impossible. But – and this is crucial – only if it respects the type of parasocial relationship you’ve already established with your human voice.
The shows that will struggle aren’t the ones that refuse to use AI. They’re the ones that use it wrong – trying to force warmth into an assistant-type show or competence into a friend-type show. It’s like wearing a tuxedo to a beach party or board shorts to a board meeting. Context isn’t just important; it’s everything. It’s like the old joke about intelligence and wisdom: Intelligence knows a tomato is a fruit, but wisdom is not putting them in a fruit salad.
Here’s my prediction, and feel free to @ me when I’m wrong: The first major podcast network to nail AI-performed host-read ads won’t do it by making the AI sound more human. They’ll do it by understanding which type of parasocial relationship each show has cultivated and matching the AI performance to that specific dynamic.
We’re not talking about replacing hosts. We’re talking about giving hosts – especially those in the struggling middle – the ability to scale their unique relationship with their audience in ways that actually generate revenue.
The question isn’t whether AI voices belong in podcasting. They’re already here, whether we like it or not. The question is whether we’ll use this research to make them work for creators and advertisers, or whether we’ll keep having the same tired debate about robots taking our jobs while the opportunity passes us by.
Because if there’s one thing I’ve learned in all my years in this industry, it’s that technology doesn’t kill media formats. Ignoring how audiences actually consume media is what kills media formats. And these researchers just handed us a roadmap for how audiences might actually accept AI in podcasting.
New Partners
Sounds Profitable exists thanks to the continued support of our amazing partners. Monthly consulting, free tickets to our quarterly events, partner-only webinars, and access to our 1,800+ person slack channel are all benefits of partnering Sounds Profitable.
- display.io: Addressable audio advertising at scale for brands and podcasters.
Want to learn more about partnership? Hit reply or send us an email!
