Audio and podcasts: ads you hear

Audio advertising reaches people when their eyes are busy: driving, cooking, running, commuting. It includes ads in music streaming, internet radio, smart speakers and podcasts, and it is increasingly bought just like display, through auctions. It also has its own quirks, because an ear cannot click and a download is not a listen.

A host-read podcast ad is like a friend recommending a restaurant at dinner. A dynamically inserted ad is more like a radio jingle that changes depending on which city you are driving through.

Two kinds of audio

Host-read vs dynamic ad insertion

Early podcasts baked ads into the audio file forever. Today most large shows use dynamic ad insertion (DAI): the episode has markers for ad breaks, and when a listener hits play or downloads, a server stitches in ads chosen for that listener's location, device and time. It is the audio equivalent of SSAI in CTV. Host-read ads can also be recorded and inserted dynamically, so they stay fresh when old episodes are replayed years later.

1. Listener requests an episodeA podcast app asks the hosting platform for the audio file.
2. Ad decisionThe host's ad server checks targeting such as country, device and show category, and may run an auction via an audio SSP.
3. StitchingAds are inserted into the break markers and the full file is delivered.
4. CountingThe download is logged and filtered following the IAB Tech Lab Podcast Measurement Technical Guidelines.

The key players

Outside the US, local services dominate: JioSaavn and others in India, regional radio groups in Europe and Latin America, and Japanese radio apps such as radiko. Many are adding programmatic audio through SSPs to reach global buyers.

How audio is bought and priced

Streaming audio is usually sold on a CPM basis, per ad played to a listener, and increasingly through private marketplace deals and open auctions in the same DSPs used for display and video. Podcasts are often sold on CPM per download, either as a sponsorship of a whole show (baked-in or host-read) or as run-of-network inventory spread across many shows by category. Host-read sponsorships usually cost more because listeners trust the host's voice. Formats are expanding too: companion display banners in the player, video podcasts on YouTube and Spotify, and short audio ads in smart speakers and in-car systems.

Measuring what you cannot click

• Downloads are the podcast currency, but a download is not a listen. Apps may auto-download episodes nobody plays.
• Streaming impressions are counted when an ad starts playing; video completion rate shows how many finished.
• Brand lift studies and lift study surveys measure recall and intent.
• Promo codes and vanity URLs (say code SUMO at checkout) are the classic podcast attribution trick.
• Pixel-based attribution matches the IP address of a download with visits to the advertiser's site, but IP matching is imprecise.

Audio fraud and quality risks

Audio has fewer signals than web or apps, which gives fraud room to breathe. Known issues include bots that request episodes over and over to inflate download counts, apps that play audio ads with the volume muted or in the background, and mobile apps that declare themselves as a popular music service in the bid request, a form of app spoofing. Data center traffic requesting podcast files is a common red flag, since real people rarely listen from cloud servers.