Advertising in podcasts behaves unlike advertising in almost any other medium. The difference comes from the fact that an episode is downloaded as a file and can be played at any point afterwards.
The file leaves the publisher's control
A podcast is distributed as an audio file that listeners fetch through many different applications, and once fetched it plays without further contact.
The publisher therefore knows a download occurred but has limited visibility of what happened next, unlike a video platform that observes every second of playback.
That measurement gap shaped the entire commercial model, since advertisers had to accept indirect evidence rather than the granular reporting they were used to.
Two kinds of advertisement coexist
Baked-in advertising is recorded into the episode itself and remains in the file permanently, so a listener hearing an old episode hears the original message.
Dynamically inserted advertising is stitched in when the file is requested, allowing the message to be changed, targeted by approximate location, and removed when a campaign ends.
Publishers use both, typically reserving the read endorsement for advertisers paying most and inserting shorter spots around it.
The host's voice is the product
A host-read advertisement carries an implicit endorsement, and listeners respond to it as a recommendation from someone they have chosen to spend time with.
That is why such reads command higher rates than produced spots, and why they are written loosely so the host can deliver them in their own manner.
Disclosure requirements apply to this kind of endorsement in many jurisdictions, and the specific obligations differ from place to place and change over time.
Response is measured by attribution
Because playback cannot be observed directly, campaigns rely on distinct codes, dedicated addresses and surveys asking new customers where they heard about a product.
Those mechanisms undercount, since many listeners hear an advertisement and act on it later without using the code, so reported performance understates the real effect.
Advertisers accordingly treat podcast results as directional rather than precise, and buying decisions rest on sustained performance rather than on single campaign figures.
Back catalogue behaves like inventory
Podcast episodes continue to be downloaded long after publication, in a way that broadcast content does not.
Dynamic insertion turns that back catalogue into sellable inventory, because a show with years of archive can carry current advertising throughout it.
This is why publishers have converted older baked-in slots to insertion points, recovering value from episodes that would otherwise earn nothing further.