IAB's December 2024 creative guidance documents an NPR test that ran one message at three lengths. The 15 second cut won on brand recall. The 60 second version won on intent. Same copy, different jobs.
Most advice about writing a podcast ad script skips that. A script is not prose you write well and then drop into whatever slot you happened to buy. It is built for one length, one placement, and one host's voice. Miss those and good writing will not rescue it.
Below: the four parts, word counts for every spot length, a copyable template, and three finished scripts. For which shows to buy rather than what to say in them, see our breakdown of podcast advertising examples.
What goes into a podcast ad script? Four parts: a hook that does not name the brand, one problem the listener recognises, two or three concrete benefits, and a call to action repeated twice. A 60 second read runs 130 to 160 words.
1. What goes into a podcast ad script?
Four parts, in this order, whatever the category. Each one has a job. Each one fails in a way you can predict.
The hook
Do not open with your brand name. A listener who hears the shape of a sponsor read reaches for skip, and your offer never arrives. Open on the problem instead, or on a fact odd enough that it still sounds like the host talking.
Here is the test. Read your first two sentences out loud. If they could pass for something the host said before the break, the hook works.
Stew Redwine of Oxford Road opened a script for Sounds Profitable with a joke about how long audio advertising has existed. It landed on the question that audience worries about: "Do the ads work?" No brand had been named. The product then arrived as the answer.
The problem frame
Narrow to one pain the listener actually lives with. The common mistake is describing it in your language rather than theirs. A meal kit brand should not say customers face weeknight meal planning friction. It should say what a tired person actually says at seven on a Wednesday night.
Tie that pain to whatever this particular audience already cares about. A finance show and a parenting show can carry the same product perfectly well. The framing should not match.
The benefit block
Two or three concrete benefits beat a feature list. Listeners cannot re read a sentence they missed, so every extra message costs you the one before it. Three is the ceiling.
- Pick one hero statement. Adam McNeil, VP of Marketing at ADOPTER Media, told Gumball an ad should centre on a single line: "broad, jargon-free, memorable, short, inspiring, and about the customer".
- Make the listener the protagonist. Your brand supplies a tool. The listener solves the problem. Reverse those roles and the copy reads as a press release.
- Attach one detail per benefit. A number, a delivery window, or a real customer outcome will outperform any adjective you reach for.
- Cut anything a host would not say. If the phrase needs a run up, it will get paraphrased on air anyway.
The call to action
The CTA has to survive driving, running, cooking and commuting. State the offer, give the URL and code, then say both again. Spell out anything that could be misheard.
Go to brand dot com and use code NAME for the offer. That is code NAME at brand dot com. Short, structured, repeated.
Write the hero statement before anything else. If it needs a subordinate clause, the campaign has no message yet, and structure will not hide that. Ten minutes on that one sentence saves an hour on the rest.
2. How long should a podcast ad script be?
Length is set by what you bought. What you control is whether the copy fits it. Voice123 puts podcast reading pace at 130 to 160 words per minute, which is the arithmetic behind this table.
| Spot length | Word count | What fits | What goes first |
|---|---|---|---|
| 15 seconds | 33 to 40 words | Brand, one benefit, CTA | The hook and the problem frame. Lead with the brand instead. |
| 30 seconds | 65 to 80 words | Short hook, one benefit, CTA with the code repeated | The host anecdote and your second benefit. |
| 60 seconds | 130 to 160 words | All four parts, two or three benefits, room for an anecdote | Nothing, if the copy is disciplined. |
| 90 seconds | 195 to 240 words | Everything above plus a longer story from the host | Nothing. Guard against padding instead. |
Shorter is not weaker. In the NPR test IAB documented, the 15 second version led on brand recall. The 60 second version drove the strongest lift in intent. Both beat the assumption that longer is safer.
These lengths do different jobs. Buying only 60 second mid rolls leaves recall on the table. Buying only 15 second spots leaves intent there instead, so a campaign that mixes them usually beats one that does not.
Word count is an estimate. Read the script aloud at the speed the host actually talks, then time it before the brief goes out. Fast hosts finish early and improvise to fill, and improvised filler is where off message copy comes from.
3. Host read or announcer read: pick one
These are two different documents. A host read is delivered by the show's own host in their words. An announcer read, also called a producer read, is pre recorded by a voice actor and dropped into the episode.
| Question | Host read | Announcer read |
|---|---|---|
| What you write | Talking points. Supply 4 to 5, require 2 to 3 per read. | Exact copy, timed to the second. |
| Who controls wording | The host, inside your guardrails | You, completely |
| Scales across shows | No. Every host needs their own brief. | Yes. One recording runs everywhere. |
| Stronger on | Purchase intent, by four points in IAB's creator economy research | Brand familiarity, by five points in the same research |
| Main risk | Host paraphrases a claim you needed stated precisely | Sounds like an interruption, not a recommendation |
Neither format wins outright. Each does different work, and the research points toward running both rather than picking a side.
The rest of this guide assumes a host read, because that is where the script itself carries the most weight. If you are writing an announcer read, write every word and time it precisely.
Ask for the host read and the announcer read in the same buy. You get one asset that converts and one that scales, and you learn which your category responds to before the next campaign.
4. Placement changes the copy
Most script advice quietly assumes a mid roll. Placement decides how much attention you inherit, and that changes the hook, the length and what the spot can realistically achieve.
How much runtime are you placing that spot into? Here is every English episode published in the 30 days to August 2026, sorted by length, from the MillionPodcasts database.
| Episode length | English episodes published, last 30 days | Same, on shows with 10K to 50K monthly listeners |
|---|---|---|
| Less than 10 minutes | 414K | 6.5K |
| 10 to 20 minutes | 119.4K | 8K |
| 20 to 40 minutes | 227.4K | 14.1K |
| 40 to 60 minutes | 174K | 14.2K |
| 60 to 90 minutes | 97.3K | 10.1K |
| Over 90 minutes | 192.1K | 22.2K |
Those bands add up to roughly 1.2 million English episodes for the month. About 75.1K of them sit on shows with 10K to 50K monthly listeners. Across all English shows, close to a third of episodes run under 10 minutes. Filter to that listener band and the share drops under 9 percent.
There the largest single band is over 90 minutes. Shows of that size skew longer than podcasting as a whole. A mid roll in a 25 minute show is not the same slot as a mid roll in a 90 minute show. Check the show's average episode length before deciding what the copy can carry.
| Placement | Attention you inherit | How the copy changes | Realistic goal |
|---|---|---|---|
| Pre roll | Low. Nobody has committed to the episode yet. | Faster hook, shorter overall, no anecdote. Reach the benefit inside two sentences. | Awareness and recall |
| Mid roll | High. The listener chose to stay through the content. | Room for a hook, a problem frame and a story from the host. | Response and conversion |
| Post roll | Lowest. Many listeners have already gone. | Shortest of the three. Offer and code, repeated, no setup. | Reinforcement for people already sold |
One consequence worth planning for. A promo code in a post roll under reports against the same code in a mid roll, every time. That gap is placement, not creative, so use a distinct code per slot if you want to read the results honestly.
Rate differences between slots are their own negotiation. Our guide to podcast ad rates covers how those conversations go.
Send three scripts, not one. Writing a separate version for each slot takes twenty minutes and removes the single most common cause of a read running long.
5. Podcast ad script template you can copy
This is a talking points brief, not a performance script. Copy it, fill the brackets, and delete whatever your spot length cannot carry.
[OPTIONAL ANECDOTE: host improvises]
Prompt: if you have used [product] or lived [the problem], open with two or three sentences here.
[INTRO: required]
[Brand] is [one sentence on what it does]. It was built to help [audience] [specific outcome].
[TALKING POINTS: host reads 2 to 3; supply 4 to 5 so reads can rotate]
1. [Benefit framed around the listener's day, not your feature set]
2. [Second benefit, plus one specific supporting detail]
3. [Social proof: a verified stat, named customer, or documented outcome]
4. [Logistics: how to get it, delivery window, availability]
5. [Price anchor or value comparison, if it helps]
[CALL TO ACTION: required, read twice]
"Go to [vanity URL] and use code [PROMONAME] for [offer]. That's code [PROMONAME] at [vanity URL]."
[TIMING]
Target 60 seconds. Scripted sections total 130 to 160 words, not counting the anecdote.
[VERBATIM FLAGS]
Mark any line that must be read word for word. Label it READ VERBATIM in bold so it survives a skim.
Supply five talking points and require two. That gap is deliberate. It lets the host rotate across a campaign. A listener who hears you three episodes running does not get the identical read.
Number the talking points and tell the host which two are compulsory. An unlabelled list of five reads as five requirements. Hosts then rush all of them, or quietly pick the two easiest to say.
A brief is scaffolding for someone else's voice, not a script for them to perform. Write tightly only where the offer or the law demands it. Everything else belongs to the host. Claim more of it and the ad stops sounding like the recommendation you paid for.
6. Three finished scripts, written out
Templates show shape. Finished copy shows judgement, so here are three at different lengths, each annotated with what the line is doing and why. Brands are illustrative; the structure is not.
Every one of these was written for a known audience. You can filter the MillionPodcasts podcast directory by beat, listener demographics and sponsor history to work out who yours is talking to.
15 seconds, 37 words, pre roll
"Notefield turns a meeting recording into a clean summary before you have left the call. No bot joins, no setup. Go to notefield dot com slash daily. That's notefield dot com slash daily."
No hook. There is no room for one at this length, so the brand leads and the benefit lands immediately. The second line kills the two objections buyers raise most. The URL still runs twice, because that rule does not bend.
30 seconds, 68 words, mid roll
"Ever bought coffee on impulse and found it three months later, still sealed? Cardinal roasts a small batch the day it ships, so it arrives tasting like someone made it for you. Pick the roast, pick how often, skip any week. Go to cardinalcoffee dot com, code SHOW, thirty percent off your first bag. That's code SHOW at cardinalcoffee dot com."
The hook is a question the listener answers in their head before the brand appears. One benefit carries the middle, with a concrete detail attached. The third line handles the objection about being locked into a subscription. Nobody explains what coffee is.
60 seconds, 136 words, mid roll
[HOST ANECDOTE: two or three sentences about the last time you hired someone.]
"Hiring for a small team is mostly rejection admin. You post a role, two hundred applications land, and your weekend goes to resumes that were never a fit.
Fieldhand is a hiring tool built for teams under fifty people. You describe the role once. It screens applicants against what you said you needed, then shows you why each one made the cut.
No applicant tracking system to configure. No per seat licence. You pay for the role you are filling, and when it closes, you stop paying.
[SOCIAL PROOF: your verified customer outcome goes here.]
Go to fieldhand dot com slash show and your first role is free. That's fieldhand dot com slash show."
All four parts are visible here. A hook naming the pain in the buyer's own words, a problem paragraph, three benefits, then a CTA said twice. The anecdote sits outside the word count because the host improvises it.
How a real brand structured theirs
Gumball published the actual brief Feals used across its sponsored reads. The move worth stealing is its personal experience section, written as open questions rather than lines to deliver. Hosts were asked what they thought of the packaging, or which plant they picked.
A host answering a question produces a real sentence. A host reading the packaging is beautiful and thoughtfully designed produces an advertisement, and listeners hear the difference immediately.
Send the product before the campaign starts. A host answering questions about something they have actually used writes your best copy for free, and it costs you one sample.
Build the list this script runs on
Copy only converts on a show whose listeners match the offer. Search 3M+ podcasts, filter on 17 dimensions including listener demographics, beat, location and whether a show already carries sponsors, then unlock verified host and booker emails and export the shortlist to CSV or Excel.
Start free, no card needed →7. What changes in regulated categories
Finance, health, supplements, insurance and anything claiming an outcome cannot run on talking points. The script becomes a word for word document, and the host's freedom shrinks to the parts carrying no claim.
- Write short sentences. Compliance language collapses when a host has to steer a subordinate clause on the third take.
- Mark the pause points. A disclosure read fast is a disclosure nobody parsed.
- Flag every fixed line. Label it READ VERBATIM in bold, because briefs get skimmed and that is exactly the line that gets skimmed past.
- Separate claim from colour. Give latitude on the story and none on the claim, then say which is which inside the brief.
- Get sign off before recording. Re records cost goodwill you will want later in the campaign.
The trade off is real. You lose some of the authenticity you bought a host read for. What you get back is a claim that survives review, and in these categories that is not optional.
Time verbatim copy against the low end of the range, so 130 words for a 60 rather than 160. Fixed copy has no slack to absorb an overrun, and the line that gets cut live is usually the disclosure.
Where to start
Write the hero statement, then the CTA, then fill the middle. Those two sentences carry the campaign, and drafting them first stops the middle from wandering.
Running one show? Write one brief, specific to that host's audience. Running twenty? Write the talking points once, then rewrite the problem frame per show, because that is where a generic script gives itself away.
The question to carry into the next brief is not whether the copy is good. It is whether the host could say it out loud without flinching.
8. Podcast ad script FAQ
How long should a podcast ad be?
Most host read podcast ads run 15, 30 or 60 seconds, and the length is set by what you bought rather than by the copy. In an NPR test documented by IAB in December 2024, 15 second spots performed best on brand recall while 60 second spots drove the strongest intent lift, so mixing lengths across a campaign tends to beat committing to one.
How many words is a 30 second podcast ad?
Roughly 65 to 80 words. Voice123 puts podcast reading pace at 130 to 160 words per minute, which works out to about 33 to 40 words for a 15 second spot and 130 to 160 for a 60 second read.
What is a host read podcast ad?
A host read ad is delivered by the show's own host in their own voice, usually from talking points rather than a word for word script. An announcer read, sometimes called a producer read, is pre recorded by a voice actor and inserted into the episode.
Should a host read script be word for word?
Not for a host read. Talking points let the host sound like themselves, which is the reason you bought a host read at all. Write word for word only where the law requires it, and mark those lines clearly so nobody paraphrases them.
Where should the promo code go in a podcast ad script?
At the very end, and twice. State the offer, give the URL and code, then repeat both in a short closing line. People listen while driving, cooking and commuting, so a code said once is a code many listeners never register.
Do I need a different script for pre roll and mid roll?
Yes. Pre roll runs before the listener is invested, so it needs a faster hook and usually runs shorter. Mid roll arrives once attention is established and can carry a longer setup plus a personal story from the host.
References
IAB. (December 2024). Podcast Creative Best Practices. https://www.iab.com/wp-content/uploads/2024/12/IAB_Podcast_Creative_Best_Practices_December_2024.pdf IAB. (August 2024). Podcasting in the Creator Economy: A Long Term Opportunity. https://www.iab.com/wp-content/uploads/2024/08/IAB_Podcasting_In_The_Creator_Economy_Guide_August_2024.pdf Gumball. (November 2022). How To Write a Host-Read Ad Script: Do's and Don'ts With Examples. https://blog.gumball.fm/how-to-write-host-read-ad-as-a-brand/ Voice123. (January 2026). Podcast Scripts: How to create awesome podcasts. https://voice123.com/blog/voice-over-scripts/podcast-scripts/ Sounds Profitable. (June 2021). How To Write Podcast Ads That Sell. https://soundsprofitable.com/article/how-to-write-podcast-ads-that-sell MillionPodcasts. (August 2026). Podcast database, episode search by episode length and monthly listeners. https://www.millionpodcasts.com/podcasts-directory