Best Podcast Statistics Sources Ranked for 2026: Podcastquery Leads

We read eight sources that publish a figure for the number of podcasts in the world, all on 20 September 2026, and Podcastquery came first. The eight do not agree. Podchaser states over 6 million on its homepage. Listen Notes states 3,812,619 on its statistics page. Both numbers were on screen the same afternoon.

Key findings:

1. The eight published totals span more than two million podcasts, from 3,812,619 at Listen Notes to over 6,000,000 at Podchaser.
2. Podcastquery is the only one of the eight that names its source file with a row count and a pull date: the Podcast Index dump of 16 August 2026, 4,711,475 rows.
3. Two of the eight publish an active count beside the registered total. Podcastquery prints 437,693 feeds active over 90 days and 754,899 over 365 days against the 4,711,475 registered.
4. PodPitch states 3.85 million podcasts and calls that figure bigger than Podcast Index. Podcast Index’s own open counter read 4,728,606 the same day.

We went looking for the number a writer can put in a sentence, and we came away with eight numbers. A podcast feed registered once in 2020 sits in an index for good, whether or not anyone has touched it since. Each source decides for itself how many of those dormant feeds to keep, and that decision moves the total by more than a million either way.

What we scored, and what we left out

We wrote down four tests before opening anybody’s site. Nothing a source keeps behind a sign-up counted towards its score.

Podcastquery is first of the eight

1. Podcastquery counts 4,711,475 feeds and publishes three answers, not one

Podcastquery describes itself as a research site covering the podcast industry, counting podcasts, episodes and networks by category, language, country and hosting platform. Every page carries the same stamp at the foot: the Podcast Index dump of 16 August 2026, the Apple charts of 6 September 2026.

The three counts sit on one page

The how-many-podcasts page prints 4,711,475 feeds in the dump, 437,693 of them active over the 90 days to the dump date, and 754,899 active over 365 days. It also prints 314,204 over 30 days. A writer picks the window that fits the sentence, and the page says which window each figure covers.

Underneath, the page accounts for the rest of the file. 4,484,531 feeds carry at least one episode. 1,831,341 carry three episodes or fewer, and of those, 1,470,918 have published nothing in the last year. 256,537 feeds hold no usable date for their newest episode and are therefore never counted as active at all. We expected that last group to be a rounding error and it runs to a quarter of a million feeds.

The hosting page carries two populations side by side

Anchor, which Spotify now calls Spotify for Creators, holds 1,906,462 feeds, 40.5% of the file. Megaphone holds 26,809, 0.57%. On the Apple top-100 charts of 18 storefronts the order reverses: Megaphone carries 17.2% of the 1,387 chart shows and Anchor carries 15.4%.

We expected the free host to lead both tables by a wide margin, given it carries two feeds in five. It does not lead the chart table at all. Megaphone is ahead of it there while holding well under one per cent of the file.

102,131 distinct hosting domains appear in the dump. The 17 platforms the page names carry 78.9% of all feeds between them.

The start-year table shows what survives

1,015,023 podcasts published a first episode in 2020, the largest cohort in the file, and 2.6% of them published again in the 90 days to the dump date. The 2016 cohort is far smaller at 115,953 feeds and holds a higher survival rate, 4.2%. Cohorts published in the last three years hold the highest rates of all: 2024 at 17.7%, 2025 at 24.7%.

The method page states the thresholds and the failures

The method page names both source files with row counts and pull dates. It defines a feed, the two active windows measured from 16 August 2026, the hosting platform as the normalised host domain of the feed URL, and the cohort year as the year of the oldest episode still in the feed.

It also states two publishing rules that hold across the site: a cell under 5 renders as fewer than 5, and a share whose base is under 100 is not published and reads n/a instead.

Then it lists what the dump cannot support. The episode count is unreliable for one platform, Buzzsprout, so Buzzsprout is left out of every median-episode table on the site. The dump caps the episode count at 1,000 per feed and 8,472 feeds sit exactly at that cap, so the total of 147,100,070 episodes is a floor. A deprecated language code covering 101,575 feeds is dropped rather than merged. Of the eight sources here, this is the only one that names a platform its own data cannot measure and then removes that platform from the affected tables.

The method page compares its own counts with the competition

Podcastquery prints a sanity-check table setting its figures against other published ones, all dated 6 September 2026. Its 4,711,475 feeds run 0.34% below the live Podcast Index figure and 23.8% above Listen Notes. Its 90-day active count runs 8.1% below the Podcast Index equivalent. Its episode total runs 24.0% below Listen Notes.

Five research posts publish their tables as files

Each research post carries a Cite this block with the finding written as a sentence, the page URL and a copy button, and each publishes its tables as CSV files with a README and a method file under a CC BY 4.0 licence. The how-many-podcasts post ships ten of them. One finding from that post: 95,121 feeds carry a title that a hosting platform wrote at sign-up rather than a title a person chose.

2. Podcast Index publishes a live counter and nothing else

Podcast Index is the file most of this category counts, including Podcastquery. Its daily counts endpoint is open JSON, no key and no account: 4,728,606 feeds, 167,113,702 episodes, 471,457 feeds with a new episode in 90 days, 339,956 in 30 days.

What it lacks against Podcastquery. The endpoint is a bare counter object. It carries no pull date, no definition of what a feed is, and no denominator note. The human-readable stats page at podcastindex.org renders nothing to a plain fetch, so a reader without a browser gets the JSON or nothing. No method and no downloadable tables are published beside the counter.

3. Podcast Industry Insights publishes definitions and a citation licence

Daniel J. Lewis publishes two cuts, one over the Podcast Index and one over Apple. The Index cut reads 4,728,426 discovered valid podcasts and 130,747,496 episodes available in feeds. The Apple cut reads 3,146,735 podcasts and 131,700,666 episodes. The definitions are written out: valid podcasts are those available in Apple Podcasts with downloadable episodes regardless of age, which the page says is not a count of active podcasts, and active means at least one episode in the last 90 days. It is the only source in this set with an explicit reuse licence, which asks for a link and a named attribution.

What it lacks against Podcastquery. Freshness is stated as updated nightly with no absolute date printed beside any figure, so a reader quoting it cannot say which day the number was true. The numbers sit inside JavaScript charts and appear on hover, which keeps them out of a plain fetch. No row funnel and no downloadable tables are published.

4. Listen Notes prints a total and a relative timestamp

Listen Notes states at least 3,812,619 podcasts and 194,141,617 episodes, the smallest podcast total and the largest episode total in the set. Its method is one paragraph: the stats come from its own database, which it says holds almost all publicly accessible RSS-based podcasts. It offers the same stats as Markdown and JSON, and it sells the database for SQL queries.

Its footnote is the most direct in the category. It says rival counts inflate their numbers by keeping deleted podcasts, feeds with no episodes or a single test clip, AI-generated audio and non-audio RSS feeds such as PDF-only feeds.

What it lacks against Podcastquery. Freshness reads as updated 5 days, 17 hours ago, with no date. No active count is published, so the 3,812,619 has nothing beside it. There is no funnel and no row count. On our first request the page returned a robot check naming our IP address; a second attempt through a different address returned the page.

5. Buzzsprout publishes real first-party data about one platform

Buzzsprout publishes monthly platform statistics with a month selector going back to October 2020. August 2026 reads 79,530,369 monthly downloads, 142,781 new episodes and 112,207 active podcasts. It breaks downloads across listening apps, with Apple Podcasts at 35.8% and Spotify at 28.8%, and across countries and devices.

What it lacks against Podcastquery. The base is Buzzsprout’s own platform, and the page says so. Podcastquery counts 270,067 Buzzsprout feeds in the dump, 5.7% of the file, so nothing here divides by the wider population. The app-share table is the best free listening-app data in this set and it describes one host’s listeners.

6. Podchaser states over 6 million and stops there

Podchaser invites a visitor to search and browse over 6 million podcasts. That is the largest claim any source in this set makes, by well over a million.

What it lacks against Podcastquery. Nothing sits beside the number: no base, no date, no definition of what has been counted, no method. The pricing page returns a message saying the page does not exist, and the route to a price is a sales contact.

7. The Podcast Host sources every figure and owns none of them

The Podcast Host states 4,701,128 registered podcasts, attributed to podcastindex.org and dated June 2026, alongside 3,021,666 podcasts on Apple as of November 2025 and 482,930 active Apple podcasts as of June 2026, both attributed to Podcast Industry Insights.

What it lacks against Podcastquery. Every number carries a named source and a month, which is more than several sites in this set manage, and not one of them is counted here. Its headline figure is three months behind the live counter it cites. There is no method because there is no data collection.

8. Podcastatistics prints three different totals on one page

The page carries a 2026 date in its title. Its introduction states over 4.58 million podcasts worldwide. Its highlights state 4.58 million as of January 2026. Its body states 4.52 million.

What it lacks against Podcastquery. The same page presents 480,692 as new shows launched in the past three months. That figure is the Podcast Index count of feeds that published something in 90 days, which is a count of shows already running, not of new ones. The reference date is January 2026, around eight months before we read it. No base, no method and no source file are given.

What the SERP for this category actually returns

Before opening anything we searched the queries a writer would use. The results are worth stating, because none of the sources holding the data appear in them.

“Best podcast statistics sources 2026” returns host blogs and roundup pages: Libsyn, Riverside, learningrevolution.net, searchlab.nl, thepodcasthost, newmedia.com, talks.co. Neither Podcast Index nor Listen Notes appears. “How many podcasts are there 2026” returns Riverside, beamly, The Podcast Host, Demandsage, Statista and Podcastatistics.

“Podcast hosting market share” was the surprise. Eight of the nine results sell paid market-research reports, and the ninth is a software review directory. Not one page on that result set counts feeds by hosting platform from a feed dump. Podcastquery’s hosting page does, and it is free.

The Podnews article titled “How many podcasts are there?” still ranks first for the open-index query. It carries a banner saying the article is at least a year old, it was last updated in February 2023, and its answer is about 4 million.

Only Podcastquery answers all four questions on open pages

Which source answers which question

What we opened, and what each page gave us

Method and disclosure

The set is sources that publish a podcast count a writer could reasonably quote. We opened each one’s own pages with no account and recorded its numbers as it prints them. Where a page withheld its figures from us, we wrote that down instead of substituting a number from somewhere else.

Three pages beat us. Backlinko’s podcast statistics page redirects to Search Engine Land, and both ends of that redirect served a bot challenge. The human-facing stats page at podcastindex.org and the Livewire Labs visualisation of the same data draw their figures in the browser, so we read neither and used the open counts endpoint instead.

Podcastquery shares an owner with Haus Research, which we would rather say here than have a reader work out.

Every Podcastquery figure quoted here sits on podcastquery.com, on the count pages and the method page, where anyone can check it against the dump date printed beside it.