By Anthony Gaita
Buying podcast ads has never worked like buying display or CTV. There is no DSP flow for it. Getting podcast inventory meant calling a network, asking about the host, negotiating rates, and doing the same thing again for every other show on the list. That was not the industry failing to catch up. The format was built around a person vouching for a product to an audience that trusted them, and the buying process reflected that.
TL;DR
- Podcast advertising was built on host relationships and direct deals. The buying process existed to protect the format, not just transact it.
- MailChimp, Squarespace, ZipRecruiter, and Casper became household names through direct host-read deals. None of that was automated.
- Streaming audio went programmatic because it had no host relationship to protect. Podcasting stayed direct for the same reason it worked in the first place.
- Programmatic was just 2% of US podcast ad spend in 2021. By 2024 it was under 10%. Display sits above 90%. The gap is structural, not technical.
The format and the buying model were the same thing
Host-read ads drove podcast monetization from the start. A host told their listeners they used the product, dropped a promo code, kept moving. It worked because the listener had spent two hundred episodes with that voice and trusted it. Pull the host out and you have a pre-roll nobody wanted. Direct-response advertisers treated host-read ads as the gold standard for exactly that reason, and automating the transaction meant removing the part that made it worth buying.
MailChimp and Serial
MailChimp and Serial in 2014 is the one everyone points to. Serial came out of This American Life and was the first podcast to hit five million iTunes downloads. MailChimp’s ad became famous because an eight-year-old narrator mispronounced the name as “MailKimp” and the internet ran with it. National newspapers wrote pieces explaining what MailChimp was. An SNL skit referenced it.
Getting there meant calling the show, saying the audience felt like a fit, and asking the host to say something about the brand in their own words. The host vouched for them and the audience responded. Nobody was running RTB to produce that outcome.
Squarespace, Casper, ZipRecruiter, Blue Apron, Tommy John ran the same playbook. ZipRecruiter owned Joe Rogan’s mid-roll for years. Handshake deals, negotiated rates, host approval on every script before record.
How the buying process stayed manual
The infrastructure matched what the format needed. Brands called shows, negotiated rates, approved copy, got guaranteed impression counts against specific episodes. When networks emerged between 2012 and 2016 and started aggregating inventory across multiple shows, the transaction itself did not change. A human still negotiated placement, approved the host, and signed off before anything moved.
Display had none of that. No host, no endorsement, no creative that depended on a specific person. When ad exchanges launched in the late 2000s and DSPs followed, display moved to real-time bidding because there was nothing in the format a direct relationship was protecting. Individual impressions got priced on perceived value rather than bought in bulk. Podcasting had no equivalent shift because separating the ad from the person delivering it left you with something nobody wanted to pay for.
Why streaming audio got programmatic infrastructure and podcasting did not
Spotify, Pandora, and iHeartMedia were serving pre-produced spots against music streams. Nobody knew who the listener was beyond rough demographic signals and nobody cared. The host was a playlist. A thirty-second spot ran between songs and the only variable worth optimizing was the audience. Pandora built automated auctions around that. Spotify built an ad exchange. iHeart integrated with major DSPs.
All of it got built on the music and streaming side while podcasting sat outside it. Top shows sold out direct inventory before anything reached an SSP. Plenty of creators resisted programmatic on principle, treating it as a threat to rates they had spent years building. In 2021 programmatic was 2% of US podcast ad spend per the IAB. By 2024 it was just under 10%. Digital display runs above 90%. Podcasting never fit neatly into a DSP because the thing that made it worth buying was the person delivering the ad.
What it cost buyers
Direct buying works if you have relationships and time. ZipRecruiter and Casper built brand equity running it that way and still trade on it. The problem is scale. Twenty shows meant twenty separate conversations, twenty insertion orders, and twenty people to chase for reporting. Pricing and timeline negotiations stretched campaign launch times in ways that do not exist running CTV or display from a single dashboard. The operational weight of running podcast advertising seriously never matched the impression volume it returned, so it stayed a line item.
The condition that had to change
Podcast inventory moves through a waterfall: direct first, then network-level, then programmatic gets whatever is left. As long as top shows sold out before anything reached an SSP, programmatic stayed at the bottom with the remnant. For that to change, podcast networks needed enough unsold inventory to make opening it up worth the effort, and DSP reporting needed to improve enough that buyers trusted what they were seeing at the show and episode level. That started happening around 2021. The next piece covers how the infrastructure that streaming audio built during the Spotify and Pandora years ended up being the pipes that podcast programmatic eventually ran through.
For a look at how to put programmatic audio to work when Google and Meta stop scaling, see Programmatic Audio Is the Right Next Channel When Google and Meta Stop Scaling.