Only 22.9% of 34,208 active English-language video podcasts show any evidence of brand activity, according to a MillionPodcasts study circulated to media contacts on August 10, 2026, leaving roughly 26,400 shows with measurable audiences carrying no advertising at all.

The figure comes from The State of Video Podcasting, 2026, a catalogue-level analysis published by MillionPodcasts, a podcast intelligence database that the company says covers 2.8 million shows. The study restricts itself to English-language video podcasts that published at least one episode in the preceding twelve months. Dormant shows are excluded from every chart and percentage. The underlying snapshot dates to May 2026.

That scoping matters for how the headline number reads. The catalogue also excludes podcasts below roughly 500 monthly listeners, on the grounds that shows under that threshold sit beneath most platform discovery surfaces and cannot be reliably modelled. What remains, therefore, is not the full undifferentiated long tail of abandoned RSS feeds. It is a filtered set of live shows with audiences large enough to model, and more than three quarters of it shows no detectable sponsor activity.

An inventory gap with a measurement problem attached

The commercial argument embedded in the data is not that video podcasting is undersold in aggregate. It is that the mechanism buyers use to identify which shows are worth buying appears to be broken.

According to MillionPodcasts, the correlation between the number of Apple reviews a show has accumulated and its modelled monthly listenership is approximately 0.84, calculated across a sample of 280 shows on a log scale. The correlation between a show's star rating and its listenership is effectively nil. The report describes review count as functioning as close to a one-to-one proxy for audience size, and puts the reason plainly: reviews do not drive listeners, they reflect them.

Star ratings fail as a signal because of compression at the top of the scale. Across the catalogue, 87% of shows score between 4.5 and 5.0 stars, and only 1.7% sit below 4.0. A rating distribution that narrow carries almost no discriminating information. The report attributes the clustering to selection among raters: listeners who would rate a show harshly tend to stop listening rather than leave a one-star review.

The practical consequence is a valuation gap. Sponsor density in the dataset tracks audience scale rather than audience approval. Sponsored shows post a median of 10,000 modelled monthly listeners against 1,500 for non-sponsored shows, a lift the report puts at 567%, or roughly seven times. Average ratings between the two cohorts are, by the report's account, identical. "Sponsorship is best understood as a consequence of audience size, not a cause of it," the report states.

Sponsor counts follow the same shape. Among the subset of shows with detailed brand records, the typical sponsored podcast carries seven distinct brands in the observed period, while the mean sits near 15, pulled upward by a small number of large network productions running 50, 100 or several hundred brands across their archives. Shows with no sponsors median around 2,000 listeners; shows carrying 11 or more brands median around 10,000.

Where the audience actually sits

The distribution underneath these figures is severely skewed, which is the second structural finding buyers face.

A typical show in the catalogue has a modelled monthly audience of about 2,000 listeners. The arithmetic mean is around 32,000, a sixteen-fold gap. The top 10% of shows capture 85.4% of all monthly listening hours. The next 40% take roughly 13%. The bottom half of the market, some 17,000 podcasts, share approximately 1% of listening between them. Around 340 shows, the top 1% by listenership, hold 41.6% of total reach. Entry to the top decile requires clearing roughly 45,000 monthly listeners.

None of that is unique to podcasting. The report describes it as the same power law that governs video channels, streaming artists and bestseller lists. What is specific to the format is the cosmetic effect of the ratings ceiling, which makes a 2,000-listener show and a two-million-listener show look nearly identical on the metric most visible in a directory listing.

Category volume and category reach diverge as well. Business and Finance leads by show count with 6,282 entries, followed by Religion and Spirituality at 3,852, Health and Wellness at 2,479, Sports at 2,298 and News and Politics at 1,909. True Crime does not appear in the top ten by show count, yet it supplies two of the ten largest podcasts by listenership, Crime Junkie and Criminal.

Audience composition figures are modelled rather than surveyed, a distinction the report flags repeatedly. Listener-weighted across the catalogue, the income split runs Low 21%, Medium 59%, High 20%. Millennials account for 48.9% of listening, Gen X 27%, Gen Z 16% and Boomers 7%. Female listeners represent 47.3% of the weighted total.

The income finding cuts against a common planning assumption. Shows whose modelled audience is dominated by high-income listeners post a median of 2,000 monthly listeners, the same as medium-income-dominant shows. Low-income-dominant shows trail at 700. Affluent targeting, in this dataset, buys no reach advantage.

Audience definition, by contrast, does correlate with sellability. Male-leaning shows number 15,233 in the catalogue and carry sponsors 28% of the time. Female-leaning shows number 7,716 with a 21% sponsorship rate, and carry nine brands on average when sponsored, more than the seven typical of male-leaning shows. Balanced-audience shows are the largest under-monetised block: 11,248 shows, sponsored 17% of the time, roughly 1.6 times less often than male-leaning inventory. Balanced shows also reach a lower median audience, 1,500 against 2,500 for shows with a clear lean in either direction.

Length, endurance and the format the market rewards

What creators produce and what audiences reward diverge sharply in the content data.

The typical episode runs about 41 minutes, and the 20-to-40 minute bucket is the most crowded, holding roughly 41% of shows with parseable length data. Median listenership moves in the opposite direction. Shows under 20 minutes median around 1,000 monthly listeners. Shows of 90 minutes or longer median closer to 6,500. The long-form cohort is thin, about 1,256 shows, and the report cautions that length and audience size are associated without any causal claim being supportable from the data.

Catalogue depth produces a wider spread. Shows with fewer than 25 episodes sit at a typical reach of 200 monthly listeners. Shows past 500 episodes reach 30,000, a 150-fold difference, though only 8% of the catalogue ever crosses the 500-episode mark. The median show has published 125 episodes. The report's reading is that consistency compounds rather than that episode count causes growth: a show reaching 500 episodes has typically published on a stable cadence for years, long enough for search visibility and listening habit to accumulate.

Guest formats are common, running on 71.8% of shows, and are associated with a median of 2,500 listeners against 1,500 for solo formats. Network affiliation is rare, detected on 4.3% of the catalogue, a figure the report explicitly labels a floor rather than a ceiling because network branding is not consistently exposed in show-level metadata.

The buyers who are already there

Among shows where brand records are available, the advertisers appearing on the largest number of distinct podcasts are, according to MillionPodcasts, Rocket Money, Shopify, Aura Frames, Amazon and BetterHelp. The composition is almost entirely direct response. The ranking counts distinct shows rather than spend, which favours brands running broad, shallow sponsorship strategies over those buying deep on a handful of properties.

Distribution has consolidated around three destinations. According to the study, 89.8% of shows publish on Apple Podcasts, Spotify and YouTube together, and a show present on all three is 2.8 times more likely to reach the top audience tier.

At the very top of the market, that pattern hardens into a requirement. All ten of the largest shows by modelled listenership distribute on YouTube, all ten run guest formats, and all ten carry confirmed sponsor activity, against 23% across the wider catalogue. Their combined modelled audience is approximately 55 million monthly listeners, led by The Joe Rogan Experience at more than 14 million and Crime Junkie at more than 9 million. Only three of the ten are formally network-backed: Call Her Daddy under SiriusXM, The Daily under The New York Times, and The Shawn Ryan Show under Cumulus. Their median rating, 4.65, sits below the dataset average. "For a video podcast, audio-only distribution is now a structural disadvantage," the report states.

What the dataset cannot support

Several figures in the report deserve handling with care, and a few contradict each other.

The rating-to-listener correlation appears twice with different values. The engagement section reports it as -0.03; the key findings section and the accompanying summary put it at -0.06. Both round to statistically zero, so the conclusion holds, but the underlying calculation is not presented consistently.

Long-form reach is reported at two magnitudes. The content section gives 90-minute-plus shows a median close to 6,500 listeners, roughly 6.5 times short-form reach and roughly 225% above the 20-to-40 minute bucket. The key findings section puts the same cohort at a median of 18,760. Catalogue depth carries a comparable gap: 500-plus episode shows median 30,000 listeners in the body text and 23,000 in the report's self-assessment scorecard, where the multiple against sub-25-episode shows is given as 4.4 rather than 150.

Platform presence is quantified two ways as well. The distribution claim of a 2.8 times higher likelihood of reaching the top tier is a different statistic from the scorecard's assertion that triple presence reaches 1.24 times the median audience of single-platform audio shows. The two are not directly comparable, and neither is shown alongside the other.

Concentration is also described inconsistently. The top decile is credited with 85.4% of listening, and the top 1%, about 340 shows, with 41.6%. The key findings section then states that roughly half of monthly listening flows to the top 1,500 podcasts, a cohort representing about 4.4% of the catalogue. Finally, the largest format cohort, sponsored guest shows on a network, is given as 821 shows in one place and approximately 2,000 in the accompanying note.

The methodological constraints are more consequential than the arithmetic. Listener counts are modelled, not measured, and the estimation engine snaps individual shows to common values such as 1,000, 2,500, 9,000, 35,000 and 100,000 rather than producing a continuous distribution. Each listener figure is therefore a band. Demographic splits are modelled from content classification, host attributes, review patterns and category priors, not from survey panels. Country fields describe where a show is produced, not where its audience lives. Revenue, advertising rates and CPM data are outside the report's scope entirely, which means the unsold-inventory finding is a count of shows, not an estimate of money left on the table.

Geographic coverage is partial. Country metadata exists for about 21,000 of the 34,000 shows. Within that subset, 79% are produced in the United States, 9% in the United Kingdom, 5.5% in Canada and 4% in Australia. Translated to the full catalogue, that is roughly 48% confirmed United States origin with another 38% unrecorded.

Host gender classification carries a similar hole. Across the full catalogue, 36.5% of shows are male-only, 16.6% female-only and 6.7% mixed, with 40.2% unclassified. Restricted to the 20,444 shows where host gender is identifiable, the splits become 61.1%, 27.8% and 11.2%. The report presents both bases rather than choosing one, and elsewhere declines to publish a median-listener premium for mixed-host shows because the underlying data block does not contain the figure, despite a headline elsewhere in the same document citing a 24% premium.

MillionPodcasts sells access to the intelligence database from which these figures are derived. The finding that reviews outperform ratings as an audience proxy, and that most of the catalogue is unpriced, is a finding that supports demand for catalogue-level intelligence products.

Why the numbers land where they do

The study arrives against a channel that has been adding spend faster than it has been adding confidence.

United States podcast advertising revenue reached 2.9 billion dollars in 2025, according to the IAB and PwC annual report. Global spending set a single-month record of 408 million dollars in December 2025. Yet podcast adoption among marketing organisations sits at 44.9%, with 30.1% reporting no plans for the channel, and only 13% of advertisers say they confidently use audio attribution tools.

Video sits at the centre of that tension. Magellan AI reported in its first quarterly measurement benchmark that podcasts distributed with video achieved a 2.49% response rate against 1.39% for RSS-only audio shows, with purchase conversion across all measured campaigns at 5.22%. Format performance, in other words, has been quantified. Which specific video shows to buy has not.

Infrastructure has been moving toward that gap from several directions. Spotify redefined a podcast play as a 30-second listen or view in June 2026 and shipped five analytics tools alongside it. IAB Tech Lab opened version 2.3 of its Podcast Technical Measurement Guidelines for comment on July 21, 2026, extending download and ad-delivery counting rules to video episodes distributed through open RSS feeds, with the comment window closing on August 19, 2026. Bumper opened its analytics dashboard to all podcasters in June 2026. Each addresses delivery, plays or first-party performance. None addresses the discovery problem the MillionPodcasts data describes, which is how a buyer sorts 34,000 shows before a campaign exists to measure.

The audience side of the equation is not in doubt. Edison Research put monthly podcast reach at 58% of Americans aged 12 and older in the Infinite Dial 2026. In the United Kingdom, Sounds Profitable found podcast advertising reaching 43% of adults monthly, with 44% reporting a purchase after hearing an ad, while the channel still captures under 5% of UK audio ad spend. The same organisation's Podcast Atlas research found 73% of listeners willing to follow creators across formats, reinforcing a creator-centric rather than platform-centric buying logic.

Monetisation economics have lagged the audience shift. Audioboom reported audio revenue per thousand of roughly 71 dollars against sub-35-dollar video rates in January 2026, a spread that limits how quickly video inventory converts into revenue even where audiences exist. On the supply side, Acast reported 26% higher revenue per listen while listens rose only 2%, a yield pattern consistent with concentrated demand chasing a narrow set of proven shows rather than expanding into the catalogue.

That is the shape the MillionPodcasts data describes from the opposite end. Roughly 4,460 shows occupy the 10,000-to-100,000 listener band, too large to run as hobbies and, on the report's reading, too small to attract major brand budgets. The report characterises this middle as the segment most exposed to the next round of network consolidation, and closes with an assessment that the format has stopped being a frontier: "The market has matured."

For media buyers, the operative question is whether a catalogue-level signal such as review volume can be built into planning workflows that currently run on downloads, plays and post-campaign attribution. For publishers outside the top decile, the finding that ratings carry no information about reach removes one of the few metrics a small show could previously point to in a pitch.

Timeline

Summary

Who - MillionPodcasts, a podcast intelligence company founded by Anuj Agarwal, which operates a database the company says covers 2.8 million shows. The findings concern media buyers, podcast networks, hosting platforms and independent creators.

What - The State of Video Podcasting, 2026, a catalogue-level analysis of 34,208 active English-language video podcasts finding that 22.9% carry any detectable sponsor activity, that review volume correlates with modelled listenership at approximately 0.84 while star ratings correlate at roughly zero, that 85.4% of monthly listening flows to the top 10% of shows, and that 89.8% of shows publish across Apple Podcasts, Spotify and YouTube simultaneously.

When - The data snapshot dates to May 2026. The study was circulated to media contacts on Monday, August 10, 2026.

Where - The catalogue covers English-language shows only. Among the 21,000 shows with recorded country metadata, 79% are produced in the United States, 9% in the United Kingdom, 5.5% in Canada and 4% in Australia.

Why - Podcast advertising spending has grown faster than the tooling used to select inventory. With United States podcast ad revenue at 2.9 billion dollars for 2025 and marketer adoption at 44.9%, the finding that more than 26,000 measurable video shows carry no advertising describes a supply pool that existing buying signals do not currently rank. The report's central claim is that the most widely visible show-level metric, star rating, contains almost no information about audience size, while a directly observable one, review count, tracks it closely.