What podcast advertising costs: the CPM arithmetic, the three slots, and what a network adds

Podcast advertising is sold on CPM, the cost of reaching a thousand downloads. The rate reports published by podcast ad marketplaces have put the average 30 second pre roll near 18 dollars CPM and the 60 second mid roll near 25 dollars, so a show averaging 10,000 downloads an episode asks roughly 180 dollars for a pre roll and 250 dollars for a mid roll. Independent shows selling a host read endorsement price above that. Large networks selling inserted spots across hundreds of shows sell below it at volume. Every quote you are sent is that one multiplication, plus a short list of extras.

So the real question is the show's download number, the slot, who reads the words, and what is added on top.

The arithmetic behind every quote

One formula produces the price of a single slot on a single episode: CPM multiplied by downloads divided by 1,000. At the marketplace averages above, the same two rates produce very different invoices as the audience grows.

Average downloads per episode	Pre roll at an 18 dollar CPM	Mid roll at a 25 dollar CPM
2,000	36 dollars	50 dollars
10,000	180 dollars	250 dollars
50,000	900 dollars	1,250 dollars
250,000	4,500 dollars	6,250 dollars

The money therefore goes on a run of episodes rather than one. A flat fee quote is the same sum backwards: divide the fee by the show's downloads and multiply by 1,000 for the implied CPM, then compare shows on that rather than on the fee.

Pre roll, mid roll and post roll

An episode carries up to three ad positions, priced by how each one is heard.

Slot	Where it sits	Usual length	Why it prices where it does
Pre roll	The first minute	15 to 30 seconds	Everybody who presses play is there, and some skip at once. The cheaper main slot
Mid roll	A break inside the episode	30 to 60 seconds	The listener has committed. The highest rate on most cards
Post roll	After the content ends	15 to 30 seconds	The smallest audience, priced that way

A mid roll on a show with half the downloads often beats a pre roll on the larger one, because you are buying attention rather than impressions. Ask where the break falls: a mid roll at minute three of a ninety minute episode is a pre roll with a mid roll price.

Host read, produced spot and programmatic

Who speaks the words changes the price more than the slot does.

• A host read endorsement. You send a brief and talking points, the host writes it in their own words and reads it. It carries the host's credibility, draws the strongest response, and is the dearest per thousand. Sought after shows charge a clear premium for it.
• A produced spot. You supply finished audio with your own voice and music. Identical on every show, cheaper per slot, and it sounds like an advert, which some categories want.
• A dynamically inserted spot. Served at the moment of download rather than edited into the file, so it can be targeted by city, swapped mid campaign and bought across a network in one order. The cheapest per thousand, with the least endorsement in it.
• A baked in spot. Edited into the file and left there, so the back catalogue keeps delivering for years after the flight ends. Count that when a show offers baked in and inserted at one rate.

This is why a host read on a small show can beat a network buy on a large one. Take a niche show with 3,000 downloads where nearly every listener does the job you sell to, at an effective 40 dollar CPM: 120 dollars an episode. The same 120 dollars buys about 6,000 inserted impressions at a 20 dollar network CPM, spread across people who mostly cannot buy from you. Twice the impressions, a fraction of the qualified attention, and the small show will sell you exclusivity too.

What a network, a marketplace or an agency adds to the price

You can buy direct from a show by emailing the host. Most spend goes through one of three intermediaries instead, and each charges for something different.

• A network or sales representative. Sells for the shows it represents and keeps a share of the ad revenue, commonly about a third. That buys ad serving, trafficking, a delivery commitment with make goods when an episode undershoots, brand safety screening, and one report across the whole order.
• A marketplace. Self serve booking across many independent shows, with visible rates and lower minimums. The right first stop for a small test budget.
• An agency. Builds the show list, negotiates, writes the brief, sets up measurement and reads the result, for a percentage of spend or a fee on top of the media.

Quoted separately from the slot: audio production, exclusivity, use of the recording off the podcast, and a rush brief. Every network also sets a minimum campaign size, so ask for that minimum in the first email. It decides whether the channel is open to you this quarter at all.

The numbers to get before you agree a price

A rate means nothing without an audited audience figure. Ask for these in writing first.

• Average downloads per episode across the last five to ten episodes, measured in the first 30 days after release. Not the best episode, and not downloads since launch.
• Whose measurement produced that number. The IAB Tech Lab podcast measurement guidelines are the standard: a download counts when a real request delivers at least a minute of audio, with duplicates from one listener removed. A show measured that way and a show quoting raw server hits are not selling the same unit.
• The audience split by country and city, which decides whether a national rate is worth paying.
• Whether the slot is baked in or inserted, the flight window, category exclusivity, and the make good policy if delivery falls short.

The rates here are the published marketplace averages and the arithmetic underneath them, written on 15 October 2026. A single show prices on its own demand: business and finance sit above general entertainment at the same download level, and a sold out show charges what a sold out show charges. Use the figures to spot a quote that is out of line, then ask why.

Measuring a channel with almost no clicks

Nobody taps an audio ad. Four methods are used in practice, and a serious buy uses more than one.

• A vanity URL and a promo code per show. Both under report, because a listener who remembers the name types it into Google instead, so read them as a floor.
• Pixel based attribution. A pixel on your site matched against devices that downloaded the episode. Spotify Ad Analytics and Podscribe both work this way, and a network will usually set one up on request.
• A post purchase survey. One question at checkout asking where the buyer heard about you. The cheapest instrument here, and the only one that still works when the ad ran six months ago.
• A lift reading. Record branded search volume, direct traffic and new customers for four weeks before the flight, then read the same three numbers during it and for a month after. Baked in reads keep delivering, so the tail is part of the result.

Set the measurement up before the first episode runs, or the flight gets renewed on a feeling and cancelled on one.

Five things buyers ask about podcast ad prices

What is a normal CPM for a podcast ad?

Marketplace rate reports put the averages near 18 dollars for a 30 second pre roll and 25 dollars for a 60 second mid roll. Host reads run above that, volume network buys below it.

Is it cheaper to buy direct from a show?

Usually, because no representation cut comes out of the rate. You take on the work the network did: finding shows, checking audience numbers, trafficking audio, chasing the make good. Direct suits a short list you know. A network suits thirty you do not.

How many episodes should a first test cover?

Enough for frequency on a small number of shows, which means a four to eight week run on two or three shows rather than one episode on ten. One episode tells you the audio played.

Why do two shows with the same downloads charge different prices?

Category, read format and scarcity. A thousand chief financial officers is worth more than a thousand general listeners, a host read is worth more than a produced spot, and a show that sells out every slot has no reason to discount. Per thousand downloads a bigger show is usually cheaper. Per qualified buyer it is often the other way round, and that is the calculation that should decide the spend.

What a buyer hears when they ask a machine what this costs

Rate cards in this channel are mostly private, so the figure a buyer walks in with came from somewhere else. Ask ChatGPT, Gemini or Perplexity what podcast advertising costs, or which network to use, and back comes a confident CPM with a short list of company names, built out of whichever pages those systems happened to read. For a network, an agency or a show selling its own inventory that list is now part of the pitch, because a buyer who never saw your rate card has already been handed somebody else's. Reading that answer for your category is the job we built AI Knows Us to do.