Who actually advertises on podcasts

We scanned the ad reads in every episode we have transcribed and counted which brands turn up, and on which shows.

Podcast advertising is famously hard to see from the outside. Networks do not publish their sponsor lists, and the ads are read aloud rather than served as tracked impressions, so there is nothing for a crawler to count. But we already transcribe every episode in order to measure talk time, and stripping the sponsor reads out is part of that. Rather than throw those segments away, we count them.

Across 915 episodes we identified 393 distinct advertisers. Here are the ones that turn up most.

Volume is not the same as reach

The two right-hand columns tell different stories, and the gap between them is the interesting part. Apple Card leads on raw volume with 259 reads, but a large share of those come from a single heavily sponsored show. Betterhelp, by contrast, appears across 12 different podcasts.

That is the difference between buying a show and buying an audience. A brand concentrated on one podcast is betting on that host's endorsement; a brand spread across a dozen is buying the medium itself and does not much mind whose voice reads the copy. Both strategies are visible in the table, and neither is obvious from listening to any single episode.

How the counting works

Sponsor blocks are found by what is said rather than where it sits, which is what lets the same method catch pre-roll, mid-roll, post-roll and dynamically inserted ads. Brand names are matched inside those blocks. One read counts once, however many times the brand is repeated within it.

The undercount is real: a novel sponsor read in an unfamiliar format can slip past the detector, and programmatically inserted ads vary between listeners, so what we transcribed is what one download contained rather than what everyone heard. Treat these as a solid floor rather than a complete census. The methodology page covers the detection in more detail.