What Counts as a Podcast Download (IAB Definition)
A podcast download is a single, de-duplicated request that pulls at least 60 seconds of an episode's audio within a 24-hour window, the definition set by the IAB Tech Lab Podcast Measurement Technical Guidelines. It is a server-log event, not a person pressing play. One download means one device asked for enough of your file to plausibly have listened. It does not mean anyone did.
That gap, between a counted request and a human listen, is the most misunderstood thing in podcasting. Get it straight and every other number on your dashboard starts making sense.
What counts as a podcast download
Under the IAB standard, a request is only counted when it clears three filters. It has to deliver a minimum amount of the file (at least 60 seconds' worth of bytes), it has to be unique within a 24-hour window for that listener, and it cannot come from a bot, a data center, or a pre-load. Strip those out and what remains is a "download."
The 60-second floor exists because anything smaller is unlikely to be a real listen, too little of the file arrived for a person to hear anything. The IAB picked one minute as a conservative threshold, in line with other ad-measured media (IAB v2.1, March 2021). The 24-hour window stops the same person from being counted twice: if you start an episode, pause it, and resume it three hours later, that is one download, not two.
Why a download is not a listen
This is the part that trips up almost everyone. A download tells you a file was requested and partially or fully transferred. It says nothing about whether a human pressed play, listened past the first minute, or finished. Auto-downloads make this concrete: many apps fetch new episodes overnight for subscribers who may never open them. Each of those still counts, as long as 60 seconds of the file moved.
So treat your download number as reach, not engagement. It is the closest the open podcast ecosystem has to an audience-size metric, because podcasts are delivered as files over RSS rather than streamed over an open connection the way YouTube or Spotify-native video is. There is no server keeping a live tally of how long each person watched. Measurement happens after the fact, from server logs, which is exactly why the IAB rules are about requests and byte ranges, not playback.
If you want true listen-through, how far into the episode people get, where they drop off, you need the analytics that hosting platforms layer on top, or Apple's and Spotify's in-app data. The download alone can't give you that.
How downloads are counted: the filter
The cleanest way to picture it is a funnel. Raw requests hit your hosting server, and the measurement layer subtracts everything that isn't a valid, unique human-scale request.
A few specifics worth knowing about each stage:
- Bots and data centers get removed. Measurement tools check requests against the IAB/ABC International Spider & Bots List and the TAG data-center IP list, so a crawler or a caching server doesn't inflate your count (Acast, IAB 2.2).
- Pre-loads are filtered. When a player or webpage silently fetches a file the user never chose, the 60-second floor is the safeguard that keeps it from counting.
- Uniqueness is per IP plus user agent. Libsyn, for example, defines a unique listener as a combination of IP address and the full user-agent string, with at least one minute of playable content downloaded (Libsyn).
- Apple Watch traffic is excluded. Because the watch and its paired iPhone request the file separately, an episode can look like two downloads from one person. The IAB has told certified companies to drop Apple Podcasts watchOS requests up front since 2020; the 2024 v2.2 guidelines reaffirmed it after members confirmed the duplication was still material.
The practical result: your IAB-certified download number is almost always lower than a raw "unique" count, because it throws away more. That conservatism is the point, it is what makes the number trustworthy enough to sell ads against.
Fixed vs rolling 24-hour windows
The 24-hour window comes in two flavors, and the difference matters more over time than on any single day. A fixed window is a calendar day, midnight to midnight. A rolling window counts the 24 hours from each request. Fixed windows can occasionally double-count a download that straddles midnight; rolling windows are more accurate but harder to compute.
For years this was optional. The v2.2 guidelines now require certified companies to disclose, in their document of methodology, which one they use (IAB v2.2, 2024). For most shows the gap is small day to day, but it compounds across months, which is one reason two "IAB-compliant" hosts can report slightly different numbers for the same feed.
How downloads relate to the rest of your feed
A download is the measured outcome of your feed working correctly. The request itself is for the audio file referenced in your enclosure tag, delivered through your RSS feed. The metadata wrapped around that file, your ID3 tags and the episode's GUID, is what apps use to recognize the episode and avoid re-fetching it, which feeds directly into the de-duplication step. Get the plumbing wrong and your counts get noisy; get it right and the IAB filter does its job cleanly.
What a "good" download number looks like
Once you know a download is reach, the obvious next question is how much reach is normal. The most-cited public benchmark is Buzzsprout's, measured in the first seven days after release (Buzzsprout global stats).
| Percentile | Downloads in first 7 days | What it signals |
|---|---|---|
| Top 50% | 28+ | You're above the median episode |
| Top 25% | 101+ | A healthy small show |
| Top 10% | 413+ | Strong, consistent reach |
| Top 5% | 1,012+ | A standout independent show |
| Top 1% | 4,611+ | Rare air |
One caveat most articles bury: these benchmarks skew indie. Buzzsprout hosts only a slice of all podcasts, and the largest hosts, Spotify/Anchor among them, publish no comparable public benchmark. So read those percentiles as a directional gauge for independent shows, not the whole market.
Frequently asked questions
Is a podcast download the same as a listen? No. A download is a server-side request that transferred at least 60 seconds of the file within a 24-hour window. It is a measure of reach. It does not confirm a human pressed play, listened past the first minute, or finished. Auto-downloaded episodes that are never opened still count.
Why is my IAB download number lower than my other stats? Because the IAB filter is stricter. It strips out bots, data-center and caching traffic, pre-loads, transfers under 60 seconds, and duplicate requests inside a 24-hour window. A raw or "unique" count keeps more of those, so it reads higher. The lower, filtered number is the one advertisers trust (Libsyn).
How long does someone have to listen for a download to count? At least 60 seconds of the file has to transfer, measured by byte range, not playback time. The IAB chose one minute as a conservative floor because anything smaller is too little of the episode for a real listen (IAB v2.1).
Do auto-downloads count as downloads? Yes, as long as 60 seconds of the file actually transfers and the request clears the bot and pre-load filters. This is a big reason download counts overstate active listening, a subscriber's app can fetch every episode automatically whether or not they open it.
What's the difference between v2.1 and v2.2 of the IAB standard? The core definition, 60 seconds, one 24-hour window, filtered traffic, is the same. The 2024 v2.2 update mainly tightened disclosure and edge-case rules: companies must now state whether they use a fixed or rolling 24-hour window, and it reaffirmed the Apple Watch watchOS filter that the IAB had recommended since 2020 (IAB v2.2).