Acast CEO Greg Glenday discusses how the podcast advertising platform is using AI to transform ad buying and optimisation – while maintaining the authentic human connections that make podcasting unique and protect it from becoming ‘AI slop’.
How do listeners feel about AI-generated podcast content?
Our research at Acast tells us that consumers don’t want synthetic content. Even if it’s a historical podcast or mini documentary series, people still have a relationship with that podcaster. Nobody wants a synthetic Marc Maron or Conan O’Brien.
When it comes to any podcast that is non-companionship – like sports scores, weather reports, traffic information – there isn’t a real relationship with a human, so we’re open to using AI there and we’re very honest with our listeners about that.
How does this thinking shape your overall AI strategy?
Podcasting has a unique market space in that it’s authentic and slower, compared with social media content for example. We call our podcasters “narrative influencers”. There’s always going to be value in advertisers investing in narrative influencers and having that word-of-mouth at scale.
I feel we have a nice moat around us where people don’t really want AI-generated content in a podcast in the same way they wouldn’t mind it in other forms of content, because podcasts are personal to the listener.
I listen to a meditation podcast where the host is very honest about their personal life, and I love that – that just wouldn’t be possible if it was AI-generated.
Maybe something like stock market reports could be AI-generated, but other than that type of content, listeners want that relationship with the host.
Can you share an example of AI-driven ‘narrative influencer’ work for a brand?
When Starbucks got a new CEO last year, we used a company we own called Podchaser to scrape 3,000 podcast transcripts for mentions of Starbucks. We put that into a large PDF showing how people were talking about Starbucks organically across major podcasts and shared it with their new CEO.
We could show them problems with the brand from the data, but also offer a solution: a narrative influencer campaign where podcasters could talk about the changes Starbucks was making. You can have a conversation with the public this way.
How are you using AI to help advertisers buy podcast ads more easily?
We have our self-serve platform, which is one of our fastest-growing platforms. We have 3,500 advertisers booking podcast ads where the average deal size is less than $10,000.
We’ve had this tool for two years, and what we found is that advertisers get overwhelmed with what to buy because we have 140,000 podcasts.
So our goal is to make podcasting easy to buy.
For self-serve, we use AI to do smart recommendations using natural language. If I owned 100 pizza places on the east coast of the US, I could put in the states they’re in, their zip codes, who their typical customer is, their price point – and ask: what podcast should I buy?
We then use our Podchaser data, which includes who is talking about pizza and your brand, and we spit out recommendations for the shows you should advertise on. We also help using AI to make the creative.
That creative can include different AI-generated accents – do you want a British accent or a Long Island accent?
How else does AI help you advise advertisers?
It helps with surfacing new shows, making the creative, building the media plan, and measuring attribution and performance.
Podcasting in the US grew on direct response advertisers. We use AI to predict when we’ll see a drop-off in attribution. We’ll proactively recommend that advertisers pause a specific show to let the audience refresh and buy other shows instead to keep attribution high. This has actually helped keep our renewal rates higher.
How is this technology helping your media planners?
Most of our business is direct, not self-serve or programmatic. But our media planners working on campaigns for big agencies started using the same self-serve platform as the SMBs because it was producing really strong recommendations – shows they might not have heard of.
That’s how we plan our direct campaigns now too. If it works for SMBs, why not use it for Omnicom, WPP, Horizon, and everyone else?
Why do you focus on selling podcasting rather than individual podcasts?
Because even big brands can reach their audience on small podcasts. We grew up selling podcasts, not podcasting. Most big advertisers put most of their money into the top 500 shows because they’re buying shows one at a time.
Most advertisers want to buy a specific podcast instead of the demographic they’re trying to reach. We put massive brands on niche podcasts that add up to the same audience size as major shows.
Are you using AI to market your podcasts today?
Our internal growth marketing team uses a lot of AI for cross-promoting shows. For example, we might put someone like Jameela Jamil, who has just launched her own podcast, onto other podcasts with a larger audience, even if they’re not as well known as her. AI helps us connect the dots internally.
How are you using AI with celebrity voices?
We use AI to recreate celebrity voices for live-read ads. Someone like Peter Crouch in the UK is very famous and difficult to get hold of, which makes live reads challenging.
It’s also useful if there’s a price change or another update to the ad – we don’t need to get them back into the studio, which would be much more expensive.
How are you using AI for content translation?
We use AI to translate some of our content. TED is one of our big, high-velocity publishers, and we produce this in Spanish now because of AI.
Is there a risk that AI will lower the quality of podcast content?
There was a shooting in Minneapolis a couple of weeks ago, and most of the content I saw online was fake and AI-generated. I wanted to find real content I could trust.
Sam Altman once said the internet is like Times Square – lots of blinking lights and people yelling. It’s chaos. The influencer and creator economy can be very fleeting.
Podcasting is different. It’s authentic and slower, and that creates long-term value.
What guardrails are you putting in place around AI use?
We’re working on our terms and conditions, and one of the things we do is inform the audience when AI has been used to create an episode – because we’re getting to the point where listeners might not be able to tell.
