Spreaker passed Anchor in 2025; 73% of its new feeds hold 3 or fewer episodes

We counted every row in the Podcast Index feed dump of 2026-08-16, 4,711,475 of them, and grouped the feeds by start year. Start year here means the year of the oldest episode still in the feed, which can be later than the year the show launched. A feed that drops its back catalogue reads as younger than it is, and we could not correct for that from this dump.

We expected Anchor to lead new feeds again. It hosts more feeds than any other platform and it led every start-year cohort from 2018 to 2024. It did not lead 2025. 106,699 feeds with a 2025 start sit on Spreaker and 86,799 on Anchor, a gap of 19,900. Spreaker was ahead of Anchor in the 2015, 2016 and 2017 cohorts, when Anchor held 1,078, 1,873 and 11,190 feeds; in 2015 Spreaker was third in the cohort, behind SoundCloud and iVoox.

Anchor's 2025 cohort of 86,799 feeds is larger than its 2024 cohort of 72,505, up 14,294 feeds, 19.7%. Spreaker's went from 53,049 to 106,699 over the same two cohorts. We cannot say whether Spreaker's 2025 feeds are the same kind of object as Anchor's, because the dump has no field separating a creator signup from a bulk-created feed. Most of Spreaker's 2025 cohort carries a Spanish language tag: 49,016 of the 60,357 Spanish-tagged 2025 feeds sit on Spreaker, 81.21%, and 42,734 of those 49,016 hold three episodes or fewer, 87.18%, with 1,107, 2.26%, publishing in the 90 days to 2026-08-16. Read the crossover with that in view.

What surprised us more was what the Spreaker feeds hold across the whole 2025 cohort: 78,261 of the 106,699, 73.35%, carry three episodes or fewer, and 56,454, 52.91%, carry one episode or none.

Key findings

1. 106,699 of the 356,901 feeds whose oldest episode falls in 2025 sit on Spreaker, 29.90%, and 86,799 sit on Anchor, 24.32%. Source: Podcastquery, September 2026.
2. 78,261 of the 106,699 Spreaker feeds in the 2025 cohort hold three episodes or fewer, 73.35%, against 20,878 of the 86,799 Anchor feeds, 24.05%.
3. 10,272 of the 106,699 Spreaker feeds published an episode in the 90 days to 2026-08-16, 9.63%, against 24,806 of the 86,799 Anchor feeds, 28.58%.
4. 49,016 of the 60,357 feeds with a 2025 start and a Spanish language tag sit on Spreaker, 81.21%, against 34,450 of the 194,338 English ones, 17.73%.
5. 45,823 of the 60,162 non-Buzzsprout Spanish 2025 feeds hold three episodes or fewer, 76.17%, against 59,738 of the 177,647 English ones, 33.63%.
6. The 2020 start cohort is the largest in the index, 1,015,023 feeds, more than the nine cohorts from 2005 to 2013 put together, 199,818 feeds.
7. 449,416 of the 997,790 non-Buzzsprout feeds in the 2020 cohort hold three episodes or fewer, 45.04%, the highest share of any cohort from 2005 to 2026.
8. 26,708 of the 1,015,023 feeds in the 2020 cohort published in the 90 days to 2026-08-16, 2.63%, the second-lowest share of any cohort in the table, behind 2011 at 2.54%.
9. Within the 2020 cohort, 62.89% of the 22,366 history feeds hold three episodes or fewer, against 29.18% of the 35,536 sports feeds.

Spreaker holds 106,699 of the 356,901 feeds that started in 2025 and Anchor holds 86,799

The 20 platforms below are the 20 with the most 2025 starts, picked by feed count alone, so the list carries three domains that are not commercial hosting platforms: loyalbooks.com, wavlake.com and xyzfm.space. Every share divides by that platform's own 2025 feeds.

Spreaker and Anchor together hold 193,498 of the 356,901 feeds in the 2025 cohort, 54.22%. RSS.com is third with 25,348 feeds, 59.99% of them at three episodes or fewer and 13.88% active. Buzzsprout is fourth with 18,217 feeds; its two episode columns print n/a because episodeCount is unreliable for Buzzsprout in this dump, with 133,259 of its 270,067 rows across all years reporting zero episodes, 49.34%. Podbean is fifth with 12,900 feeds, 14.99% at three or fewer. loyalbooks.com has fewer than five feeds at one episode or none, so that cell prints <5 in the CSV and its share prints n/a.

Hubhopper runs 86.40% of its 2025 feeds at three episodes or fewer, wavlake.com 77.72%, Spreaker 73.35% and RSS.com 59.99%. Megaphone runs 6.37%, Omny Studio 6.94% and Libsyn 8.47%. The active column runs the other way: Megaphone 51.42%, Libsyn 50.90%, Podigee 48.77%, Spreaker 9.63%.

78,261 of the 141,460 feeds in the 2025 cohort at three episodes or fewer, with Buzzsprout removed from both sides, sit on Spreaker. That is 55.32%. On the all-platform base of 143,629 the same numerator reads 54.49%.

Spreaker's 2025 cohort leads Anchor's by 19,900 feeds after seven cohorts behind

Anchor led every cohort from 2018 to 2024. In 2024 it was ahead by 19,456 feeds, 72,505 against 53,049. In 2025 Spreaker is ahead by 19,900, 106,699 against 86,799. The 2026 row covers seven and a half months, to the dump date of 2026-08-16, and both platforms are still adding to it, so we print it as partial and its 163-feed gap with it.

Anchor's new-feed counts by start year run 510,405 in 2021, 118,152 in 2022, 88,849 in 2023 and 72,505 in 2024. The drop from 2021 to 2022 is 392,253 feeds, 76.9%, and from 2022 to 2024 a further 45,647 feeds, 38.6%. Then 2025 turns up to 86,799. Anchor's total share of the index and its 2020 dominance are covered in our earlier report on Anchor's 40% of feeds and 15% of the Apple charts.

Spreaker's share at three episodes or fewer runs 12.72% in the 2022 cohort on 36,377 feeds, 32.05% in 2023 on 27,251, 59.76% in 2024 on 53,049 and 73.35% in 2025 on 106,699.

RSS.com peaked earlier. Its largest cohort is 2022 at 34,160 feeds, then 31,522 in 2023, 27,901 in 2024, 25,348 in 2025 and 15,240 in 2026 to 08-16. Its share at three episodes or fewer fell from 82.33% in 2021 to 59.99% in 2025.

81.21% of Spanish-language 2025 feeds sit on Spreaker against 17.73% of English ones

The language tag is self-declared in the feed and we did not check it against the audio. The 15 languages in the dataset cover 338,503 of the 356,901 feeds in the 2025 cohort, 94.85%.

49,016 of the 60,357 Spanish-tagged 2025 feeds sit on Spreaker, 81.21%, and 5,493, 9.10%, on Anchor. Of those 49,016 Spanish Spreaker feeds, 42,734 hold three episodes or fewer, 87.18%, and 1,107 published in the 90 days to 2026-08-16, 2.26%. English runs differently: 34,450 of the 194,338 English 2025 feeds sit on Spreaker, 17.73%, and 49,474 on Anchor, 25.46%. Of the English Spreaker feeds, 19,494 of 34,450 hold three episodes or fewer, 56.59%.

Portuguese and Italian look like Spanish. Spreaker holds 67.07% of the 10,020 Portuguese 2025 feeds and 69.76% of the 9,386 Italian ones, and 81.62% and 50.43% of its feeds in those two cohorts hold three episodes or fewer. Japanese and Vietnamese sit at the other end: Spreaker holds 1.69% and 0.53% of those cohorts, and Anchor 69.06% and 76.16%.

We could not establish why the Spanish cohort behaves this way. The dump carries the counts and nothing about where the feeds came from, and we found no statement from Spreaker on 2025 signups to compare against.

The 2020 cohort is the largest at 1,015,023 feeds and runs 44.35% at three episodes or fewer

2020 is the largest cohort at 1,015,023 feeds. 450,167 of them hold three episodes or fewer, 44.35%, and 449,416 of the 997,790 with Buzzsprout removed, 45.04%. 272,329 hold one episode or none, 26.83%. 26,708 published in the 90 days to 2026-08-16, 2.63%, the second-lowest share in the table behind 2011 at 2.54%.

The lowest share at three episodes or fewer in the table is 2005, 24.14% on 2,945 feeds, and the next lowest is 2016 at 24.60% on 115,953. 2025 runs 40.24% on 356,901 feeds, the second-highest share in the table after 2020.

Read the active column with care. It rises with recency after 2021, to 68.07% for the partial 2026 cohort. A feed whose oldest episode falls in 2026 has had at most seven and a half months since that episode. The 2026 row's 100.00% at 365 days is arithmetic, since no such feed can have a newest episode more than 365 days before 2026-08-16.

In the 2020 cohort, history runs 62.89% at three episodes or fewer and sports 29.18%

category1 in the dump is the feed's first self-declared Apple category, so this is a split of declarations.

History tops the 2020 table at 62.89% of 22,366 feeds at three episodes or fewer, and sports runs 29.18% of 35,536, 33.71 points below it. Games runs lower still, 23.29% of 425 feeds. On the active column games runs 6.35%, above religion at 5.64% and sports at 5.16%, and all three sit well above arts at 1.34%.

The 2025 cohort reorders the same categories. Society is the largest 2025 category at 90,946 feeds, 25.48% of the cohort, and carries the highest share at three episodes or fewer of any 2025 category, 63.42%. History, top of the 2020 table, runs 28.39% in 2025 on 5,723 feeds.

4,449,549 of the 4,711,475 rows carry a start year between 2005 and 2026

Three lines are dropped. 256,764 rows carry no oldest-episode date at all and so have no start year. 4,486 carry a date earlier than podcasting itself, before 2004, which we read as a placeholder. 676 fall in 2004, before the published table starts. Nothing else is dropped: 256,764 plus 4,486 plus 676 is 261,926, and 4,449,549 plus 261,926 is 4,711,475. Buzzsprout rows stay in every feed count and every active count, and come out only of episode cuts, where they are removed from numerator and denominator together.

Method

Host is the registrable domain of the feed URL, lowercased, with the alias map in tools/count.py folding domains belonging to one company. spotify.com, spotify.link, podcasters.spotify.com and creators.spotify.com fold into anchor.fm, which prints as "Anchor (Spotify for Creators)". Spreaker is spreaker.com. 26 rows carry an empty host.

Start year is the year of oldestItemPubdate, the oldest episode still in the feed. A show that trimmed its archive counts in a later cohort, which inflates the recent cohorts by an amount this dump cannot measure. That effect bears hardest on the 2025 and 2026 rows, and it is the caveat we would check first against any figure here.

A cut at three episodes or fewer counts episodeCount at three or fewer, and a cut at one episode or none counts it at one or fewer; both include feeds reporting zero. Active 90d means a newestItemPubdate inside the 90 days before 2026-08-16, active 365d the same over 365 days. Feeds with no newest-episode date are never active and stay in every denominator. episodeCount is capped at 1,000 for recently refreshed feeds, so any episode sum from this dump is a floor; the cap does not touch a cut at one or three episodes.

Buzzsprout carries no episode figure anywhere here, because 133,259 of its 270,067 rows report zero episodes. Removing Buzzsprout from both sides moves the 2020 share at three episodes or fewer from 44.35% of 1,015,023 feeds to 45.04% of 997,790.

As a sanity check, our 4,711,475 feeds sit 17,090 below the 4,728,565 reported as feedCountTotal by https://stats.podcastindex.org/daily_counts.json, read 2026-09-15, 0.36% low, which is the direction and size we expect from a dump a month older than the live counter. The same file reports 472,076 feeds with a new episode in the last 90 days against our 437,693 for the 90 days to 2026-08-16, again the expected direction.

Two limitations matter more than the rest. The first is the start year above. The second is what a feed is. A feed with three episodes may be a live show that trimmed its archive or a duplicate of another feed. A count of feeds on a platform says nothing about that platform's customers or revenue. We counted feeds holding three episodes or fewer and whether each one published inside a window, and we make no claim about why any feed holds the episodes it holds. The full write-up is on /method.

Data

Every table above is published as CSV: new_feeds_by_host_year.csv, 252 rows, one per platform and start year from 2015 to 2026; cohorts_all.csv, 22 rows, one per start year from 2005 to 2026 with and without Buzzsprout; cohort_2020_by_category.csv, 61 rows; cohort_2025_by_category.csv, 63 rows; and language_2025.csv, 15 rows. README.md names every column and denominator and METHOD.md holds the host aliases, the definitions and the limits. Cells under 5 print <5, and a share whose numerator is masked or whose base is under 100 is empty in the CSV and prints n/a in the tables above. Data: CC BY 4.0, Podcastquery, September 2026. Next to any number here, cite: Source: Podcastquery, September 2026.