Industries / Media, arts and recreation / Media and entertainment / Podcasting
Measuring the audience
Counting how many people hear or watch a show and finding out who they are, through surveys, server logs (the record a web server keeps of every file request) and the play data that apps collect.
7
Milestones achieved, from 2006 to 2025.
3
To watch for: announced, not yet confirmed outside.
3
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
2006
Surveys start counting listeners
By January 2006 Arbitron and Edison Media Research's yearly US survey of digital audio was asking about podcasting, and their 2007 report found that awareness of podcasting rose from 22% of Americans aged 12 and over in January 2006 to 37% in January 2007, while the share who had ever listened rose only from 11% to 13%, about 32 million people. The same survey, now run by Edison Research at SSRS, found in 2026 that 58% of Americans aged 12 and over had listened to or watched a podcast in the past month and 45% in the past week.
LessonA survey counts people, which a server cannot, but it depends on what people remember and report.
Credit
Arbitron; Edison Media Research (now Edison Research at SSRS)
2017
One rulebook for downloads
IAB Tech Lab's Podcast Measurement Technical Guidelines, version 1.0 in September 2016 and version 2.0 in December 2017, set out how to turn server logs into a download count. Under version 2.0 a request counts only if the file's header and about one minute of audio were sent; requests from the same internet address and the same user agent (the app name a program sends with each request) count once a day; and known bots are dropped. The document says that counting from logs "counts only the file downloads and not the actual listening." Version 2.2 (2024) added rules on invalid traffic, and version 2.3, final on 29 September 2026, applies the same rules to video files delivered through RSS feeds.
LessonA shared rule makes counts comparable between companies, but it does not make a download a listen.
Credit
IAB Tech Lab's podcast technical working group, led by Steve Mulder of NPR, with the IAB Audio Committee
2017
Apple shows what was played
On 14 December 2017 Apple released a free podcast analytics dashboard in test form. For its own app it showed unique devices, total time listened, time per device, average share of an episode played and where in an episode listeners dropped off; TechCrunch noted that podcasters had until then relied on download counts from servers. Podnews noted that it covered only Apple Podcasts on iOS 11 and iTunes 12.7, perhaps a fifth of all plays by Podnews's estimate, and that it counted devices, not people.
LessonPlay data from one app is precise but partial.
Credit
Apple
2018
Outside audits of counts
In December 2018 IAB Tech Lab began certifying companies against its guidelines, with Blubrry and NPR the first two; the audits covered code review, test measurements and log checks, and certified firms get a seal and a listing. France went further: under a November 2024 agreement, Médiamétrie and the ACPM, the French audit body, adopted one method in which the ACPM collects and checks publishers' logs, and Médiamétrie's monthly podcast figures now come from those logs filtered to IAB Tech Lab rules; by its count, 140 million podcasts were listened to or downloaded in France in October 2026.
LessonAn audit checks that the rules are applied, not why a file was fetched; in September 2022 Bloomberg, as summarised by Podnews, reported publishers buying 20-second plays inside mobile games that still counted as certified downloads because more than 60 seconds of audio was downloaded.
Credit
IAB Tech Lab; Blubrry; NPR; Médiamétrie; ACPM
2022
An open, public counter
In September 2022 John Spurlock launched OP3, the Open Podcast Prefix Project, a free counter with open code under the MIT licence. A show adds a prefix, a redirect address placed in front of each episode's file address, so every request passes through OP3, and each show gets a public stats page, with the request data available to anyone through a public API; the hosting company Transistor offered it to its shows from 19 September 2022. In May 2024 Podnews used a week of OP3 data, 3.7 million downloads across 1,458 shows with about 5.3% from bots, to show which apps people use in each country; OP3 says it now counts more than 19 million downloads a month across more than 5,400 shows, and it is not IAB-certified.
LessonOpen request data lets outsiders check and reuse a count instead of taking it on trust.
Credit
John Spurlock
2023
Apple's update shrinks downloads
Apple Podcasts pauses automatic downloads when a listener has not played a show's latest five episodes for 15 days. Before iOS 17, released on 18 September 2023, a returning listener's app fetched every unplayed episode; under iOS 17 it fetched only new ones and stopped treating republished episodes older than seven days as new. Download counts fell as phones updated: Podtrac saw a drop of about 15% for most large shows, Triton Digital found almost a third of US weekly downloads gone year on year, and Acast reported a 12% fall in listening and a $7.2 million hit, while Apple explained the change only in a short note dated 11 October 2023.
LessonA download count can fall because one app changes a setting while what people actually hear stays the same.
Credit
Apple (the change); Podtrac, Triton Digital, NPR and Podnews (measuring and reporting it)
2025
Video joins the count
In February 2025 YouTube said it had "more than 1 billion monthly active viewers of podcast content" as of January 2025, without defining a viewer; Radio Ink reported that on YouTube "any time spent listening counts as a full podcast view." Outside surveys confirmed the shift: Edison's Infinite Dial 2025 found YouTube the service US weekly podcast listeners use most (33%) and that 51% of Americans aged 12 and over had watched a podcast; Edison's quarterly top-50 ranking, a survey of 5,000 weekly listeners that measures reach, not downloads, began counting people who only watch; and Edison's July 2026 UK study put YouTube (29%) ahead of Spotify (28%) for the first time.
LessonA YouTube view, a Spotify play and a download are counted under different rules, so they cannot be added together; Edison notes that YouTube's chart "ranks by watch time rather than reach.".
Credit
YouTube (Tim Katz); Edison Research
Watch for
A play, not a download
The Alliance for Measurement in Podcasting (AMP), an industry group whose members include Spotify, Oxford Road, Libsyn and Podscribe, ratified four common definitions for audio and video, among them a "Podcast Play" of 30 seconds of content played, with the download kept alongside it as a signal of delivery. Spotify adopted a 30-second play for audio and video on 11 June 2026 and did not restate past figures.
Achieved whena platform other than Spotify, such as Apple Podcasts or YouTube, or a certified host reports Podcast Plays to the AMP definition, and an outside auditor checks one such count.
Claimed by
AMP and Spotify, June 2026
Spotify plays outlast YouTube views
Podstock, a company that matches the same episode across platforms, says a Spotify stream averages 1.5 times the viewing or listening time of a YouTube view of the same episode, based on "thousands of episodes" over 30 days, with no episode count or spread given.
Achieved whenan outside party publishes a matched-episode comparison, with its episode count and error bars, that reproduces or corrects the ratio.
Claimed by
Podstock (chief executive Michael Paretzky), May 2026
Rules for streamed video
IAB Tech Lab says version 3.0 of its guidelines, due in 2027, will focus on streaming video podcasts; version 2.3 leaves "true streaming," delivery in real time reported by the player, out of scope.
Achieved whenversion 3.0 is final and at least one company is certified against it.
Claimed by
IAB Tech Lab (Anthony Katsur), July 2026
Open
01
Test the certified counters
A test show you own (your own recordings or public-domain audio, with no ads), placed on five or more hosts that hold IAB Tech Lab certification, with the OP3 prefix added. Send each host the same scripted set of requests: full downloads, the two-byte probe Apple's app makes, partial downloads under one minute, repeats from the same internet address and app within 24 hours, the same address with different apps, and known bot names. Publish each host's reported count next to what the IAB rules say the count should be.
The briefCredit · unclaimed02
An open count by country
A public table of the share of adults in each country who listen to or watch podcasts monthly. Where a survey exists, use the published figure (Edison's Infinite Dial and UK studies, the Reuters Institute Digital News Report, Médiamétrie); where none exists, model an estimate with a stated range from open indicators such as the ITU's statistics on internet use and OP3 downloads by country.
The briefCredit · unclaimed03
Converting views, plays and downloads
A public tool and dataset that matches the same episode across YouTube, Spotify and the RSS feed and converts the three counts into one figure, such as people who played at least 30 seconds (the AMP definition), with stated error bars, fitted and tested on shows that share all three counts.
The briefCredit · unclaimed
Open 01 · the brief
Test the certified counters
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A test show you own (your own recordings or public-domain audio, with no ads), placed on five or more hosts that hold IAB Tech Lab certification, with the OP3 prefix added. Send each host the same scripted set of requests: full downloads, the two-byte probe Apple's app makes, partial downloads under one minute, repeats from the same internet address and app within 24 hours, the same address with different apps, and known bot names. Publish each host's reported count next to what the IAB rules say the count should be.
Why it is still open
The rules leave room for different results: version 2.0 allows "more sophisticated mechanisms (like a rolling 24-hour window)," version 2.2 only requires firms to declare which window they use, and certification produces a seal and a listing, not published test results. This research found explanations of why counts differ but no controlled test: LiveWire's comparison of prefix and host counts (January 2023) ran none, Podnews (May 2024) compared OP3's app shares with host data, not counts, and The Podcast Host (2025) compared Rephonic's estimates with 14 shows' host numbers but took the hosts' numbers as true.
Built with what exists
curl or Python request scripts; two or three home and mobile connections; host dashboards and CSV exports; the OP3 public API; a public GitHub repository.
First step
Make a three-episode test show, put it on two hosts with the OP3 prefix, run the 10 scenarios once and compare the counts a week later. Run no ads, so no advertiser pays for test traffic; keep volumes small, read each host's terms and tell the hosts in advance.
Done looks like
A published table of at least 5 certified hosts by 10 scripted scenarios, each run on 3 separate days, giving the expected count, each host's count and the percentage difference, with the scripts and raw request logs released.
Open 02 · the brief
An open count by country
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public table of the share of adults in each country who listen to or watch podcasts monthly. Where a survey exists, use the published figure (Edison's Infinite Dial and UK studies, the Reuters Institute Digital News Report, Médiamétrie); where none exists, model an estimate with a stated range from open indicators such as the ITU's statistics on internet use and OP3 downloads by country.
Why it is still open
The largest cross-country survey, the Reuters Institute's, covered 48 markets in 2025, only four of them in Africa. OP3 has data by country but, as Podnews notes, "isn't, yet, of the size that makes it wholly representative." This research found no open, reproducible worldwide estimate, only paywalled forecasts from eMarketer and Statista.
Built with what exists
Python and pandas; PyMC or statsmodels for estimates with ranges; ITU open data; the OP3 API.
First step
List every published national figure with its definition (monthly or weekly, age range, listen or watch, year).
Done looks like
Estimates for at least 100 countries, each marked as surveyed or modelled, with a test that hides each surveyed country in turn and reports the average error in percentage points, and code and data released under an open licence.
Open 03 · the brief
Converting views, plays and downloads
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public tool and dataset that matches the same episode across YouTube, Spotify and the RSS feed and converts the three counts into one figure, such as people who played at least 30 seconds (the AMP definition), with stated error bars, fitted and tested on shows that share all three counts.
Why it is still open
Podstock's matched comparison is commercial and publishes no episode count or spread; AMP's common unit has been adopted only by Spotify so far; IAB's version 2.3 covers only files delivered through RSS, and version 3.0 is not due until 2027. This research found no public conversion with error bars.
Built with what exists
The YouTube Analytics API, with each channel owner's permission; Spotify for Creators exports shared by participating shows; OP3 public download data; whisper.cpp (see the Note "A first stable number for local AI") to transcribe the first minutes of each version and match episodes; Python and statsmodels. Use official APIs and owners' consent only, with no scraping.
First step
Recruit five shows that already use OP3 and also publish on YouTube and Spotify, and ask each for 30-day exports of 10 episodes.
Done looks like
At least 30 shows with 10 or more episodes each and all three counts for every episode, published conversion factors with 90% ranges, and a test on 10 held-out shows that reports the error.