The 90-Minute Problem: Why the Biggest Podcasts Keep Getting Longer
Flagship shows have crept from a tidy 35 minutes to two hours and beyond. We chart the drift across a decade of feeds, ask what listeners actually finish, and name which shows earn the runway.
There is a quiet arms race in podcasting that nobody campaigns about and everyone has noticed. The shows that sit at the top of the charts in 2026 are, on average, twice the length they were in 2016. Episodes that once breezed through a commute now eat the entire dog walk, half the cooking, and trail off somewhere in the washing-up. The medium has bulged.
This is partly an artefact of who won. The most-downloaded podcasts in the English-speaking world are conversational, host-led, lightly produced, and structurally allergic to a hard out. The shows that lost — or were forced into different formats — were the heavily-edited narrative ones, which are slower to make, more expensive per minute, and held to a more disciplined runtime by the simple fact that an editor has to defend every second.
So we have ended up with a top of the charts that rewards length and a middle of the charts that cannot afford it. Whether that is good for the craft is a different question, and it is the one this piece is interested in.
How long is long now?
I pulled median episode duration for fourteen flagship shows across two snapshots — the first full year I could verify in their feed, and the trailing twelve months ending March 2026. Where a show launched after 2016, the first column is its launch year. All figures are in minutes, rounded to the nearest five.
| Show | First-year median | Trailing-12-month median | Change |
|---|---|---|---|
| Joe Rogan Experience | 130 (2016) | 175 | +35% |
| Lex Fridman | 90 (2018) | 195 | +117% |
| Tim Ferriss | 110 (2016) | 130 | +18% |
| Conan O'Brien Needs A Friend | 80 (2018) | 105 | +31% |
| SmartLess | 65 (2020) | 75 | +15% |
| Hard Fork | 60 (2022) | 80 | +33% |
| Decoder | 50 (2020) | 70 | +40% |
| Search Engine | 35 (2023) | 50 | +43% |
| The Rest Is Politics | 45 (2022) | 55 | +22% |
| The News Agents | 35 (2022) | 45 | +29% |
| The Daily | 22 (2017) | 28 | +27% |
| Today, Explained | 22 (2018) | 25 | +14% |
| 99% Invisible | 25 (2016) | 35 | +40% |
| This American Life | 60 (2016) | 60 | 0% |
A few things stand out. The conversational tier has run away with itself: Lex Fridman has more than doubled, and Joe Rogan — already long in 2016 — added a casual 45 minutes to the median. The British political duopoly (Rest Is Politics, News Agents) has crept up around 25%. Public-radio descendants have held the line; This American Life is the same length it was a decade ago because Ira Glass cuts to the runtime, not from it. The Daily has expanded by six minutes, which sounds modest until you remember it ships five times a week.
Why every show wants more minutes
The drift is not vanity. There are at least four mechanical reasons a producer chooses a longer cut, and only one of them is about the listener.
- Mid-roll inventory. The IAB threshold for a second mid-roll is 30 minutes; for a third, roughly 60. A show that crosses 60 minutes can sell three slots instead of two, and a show that crosses 90 can sell four. The revenue uplift per added quarter-hour is, in advertising terms, free money.
- Algorithmic stickiness. Spotify and Apple weight completion rate, but they also weight total minutes consumed per user per week. A 90-minute episode that 60% of subscribers finish outperforms a 35-minute episode that 95% finish, on the metric that drives recommendations.
- Production economics. A two-hour conversation costs roughly the same to record as a one-hour one. Editing is the variable cost. Lightly-edited shows have therefore drifted towards the runtime that the host's stamina, not the editor's patience, decides.
- The guest's calendar. A serious guest gives you an hour. A serious guest who is enjoying themselves gives you two. Few hosts are willing to bin the second hour once they have it.
None of these forces pull towards the listener. They pull towards the platform, the sales team, the producer's P&L, and the host's ego. Which is fine; podcasts are a business. But it is worth being honest about whose interest the runtime is serving.
What listeners actually finish
Edison Research's Infinite Dial 2026 and the most recent Podtrac subscriber panels give a fairly consistent picture: completion rate falls roughly linearly with duration once you cross 45 minutes, then slumps harder past 90. The headline numbers, smoothed across genres:
- Under 30 minutes: ~82% completion
- 30–60 minutes: ~71%
- 60–90 minutes: ~58%
- 90–120 minutes: ~44%
- Over 120 minutes: ~31%
A two-hour episode is, on average, fully consumed by less than a third of the people who pressed play. That does not make it a bad episode. It means the listener relationship with a long show is structurally different — closer to a YouTube video that gets sampled than a magazine feature that gets read end-to-end. Hosts who insist their three-hour interviews are being absorbed in full are flattering themselves.
Which long shows earn it
Not all bloat is equal. Some podcasts use length as scaffolding; others use it as filler. After a year of listening with a stopwatch in mind, here is where I land:
Earns the runtime
- Conan O'Brien Needs A Friend. Conan and his producers have a near-perfect feel for when a bit has crested. The 105-minute median feels like 70.
- Hard Fork. Casey Newton and Kevin Roose treat the conversation as a column with footnotes; their 80-minute episodes have the structure of a 1,500-word piece, with digressions earning their place because they relate back to the spine.
- Search Engine. PJ Vogt's reporting demands the extra fifteen minutes; the show's growth from 35 to 50 minutes tracks with deeper reporting, not slacker editing.
- The Rest Is History. Two genuine experts; the runtime is the format. You are not listening for compression.
Has not earned it
- Lex Fridman. A three-and-a-quarter-hour median is not a craft choice; it is a refusal to edit. The interviews would benefit from being half as long and twice as prepared.
- Joe Rogan. When the guest is good, the runtime is irrelevant. When the guest is filler, the show keeps recording anyway. The hit rate per minute has fallen as the median has risen.
- SmartLess. Pleasant, but the third act is almost always droopier than the first. A 60-minute cut would be a stronger show than the 75-minute one being released.
The distinction is craft, not genre. A long show works when the host treats the runtime as a budget to spend; it sags when the host treats it as a property to fill.
A modest argument for the 35-minute episode
The most underrated runtime in 2026 is the one This American Life has held to since the Clinton administration: somewhere between 30 and 60 minutes, with the discipline of a hard out. It is the length at which an editor still matters, at which a single act has to do real work, and at which the listener finishes. It is also the length the algorithm actively punishes, which tells you something about the algorithm rather than the form.
If you produce a podcast and you are tempted to add fifteen minutes to your median because the sponsor brief allows it or the platform rewards it, consider the inverse experiment first. Cut fifteen minutes. See whether anyone notices. The shows I listened to most often this year — The Daily, Search Engine, 99% Invisible, This American Life — are all on the short end of their genre. That is not a coincidence. It is the only honest creative constraint left in a medium where every other force pushes the timer up.
The 90-minute problem is not really about minutes. It is about whether the runtime is set by what the show needs or by what the business prefers. Listeners can tell the difference. They are voting with the skip button, and the data says they are voting often.