Podcast news roundup: 17 September 2026

Podcast Week in New York produces the research the rest of the year argues about, and the week of 10 to 16 September 2026 delivered: a listener study that puts the host at the centre of everything, and an ad test that tried to clone the host's voice. The bigger practical change for a British show came from elsewhere. Apple released iOS 27 on 14 September with a redesigned Podcasts app for Apple TV, and on 16 September Amazon Music announced video podcasts for UK customers. Add a measurement report, a quiet rule change at Spotify, and the edit desk on a question hosts rarely say out loud: how do you get a guest to do what you need?

Why does Apple putting podcasts on the television matter to your episode page?

Apple's Newsroom post of 14 September lists a redesigned Apple Podcasts app for Apple TV, and the creator support page fills in the detail. The tvOS app is completely rebuilt: it plays video podcasts, supports premium subscriptions and network channels, show pages were redesigned around show art, and episode lists now display episode art. On the Mac, video episodes are clearly labelled and pressing play goes straight into video, full screen or Picture in Picture.

Two smaller changes matter more. Video episodes across iOS, iPadOS, macOS and visionOS now include full read-along transcripts, chapters and media links. And a new Search in Show bar, on iPhone, iPad, Mac, Apple Vision Pro and the web, searches a show by guest name, episode title, keyword, or something mentioned in the transcript. Not every feature reaches every country.

My reading: the episode page has just become the thing the television shows and the thing the search bar reads. Real episode art looks like a programme on a TV; one piece of cover art repeated down the list looks like a feed. Three jobs: name the guest on mic, write chapters that describe the episode rather than decorate it, and make episode art. If titles and chapters are the part you skip, SEO titles and show notes is where we take that off your plate.

Amazon Music brings video podcasts to the UK

Amazon Music announced on 16 September that its customers can now access shows hosted on ART19, Triton Digital's Omny Studio and Flightcast, with SiriusXM (including SiriusXM Media, AdsWizz and Simplecast), Megaphone and Libsyn expected in the following months. Video podcasts are available in six countries: the US, Canada, the UK, Mexico, Brazil and Japan. Creators can dynamically insert video ads while keeping existing sponsorships and baked-in ads, and publishers keep control of hosting, ad inventory and targeting. Flightcast says hundreds of its creators will be there on day one; The Hollywood Reporter notes that ART19 is Amazon-owned.

What it means depends on your host. On Omny, ART19 or Flightcast, your video feed has a new UK destination this week. On Megaphone, Libsyn or Simplecast, you are in the queue, and "the following months" is not a date. My opinion: the interesting part is the ad model. Amazon is letting publishers keep their sponsorships, their baked-in reads and their inventory, which is now the bar for any platform that wants video from independent shows. Still deciding whether video belongs in your workflow? Start with the video podcast production guide.

What did The Podcast Study 2026 say about why people listen?

The Podcast Study 2026, presented at the Sounds Profitable Business Summit and released on 15 September by Strategic Solutions Research (Hal Rood) and Point-To-Point Marketing (Tim Bronsil), surveyed 1,205 podcast consumers in a nationally representative sample. The headline is the host: 82% say the host is a big part of, or the only reason for, listening to their favourite podcast, and 88% spend at least half their listening time with one favourite show. On host endorsements, 57% say they are more likely to remember the brand and 44% say they are more likely to trust the ad, and 45% are at least somewhat likely to pay for a tier that removes all advertising except host-read ads.

Now the companion piece. On 13 September IPVanish, the VPN owned by Ziff Davis, and Blue Wire, a network of more than 250 independent sports podcasts, published a test: same offer, identical script, three formats. Human host-reads, outsourced announcer-reads, and AI replicas of the hosts' own voices through AdBridge's Trusted Voices, measured on net-new paid subscriptions. The AI replicas beat the announcer reads by 26%. The real hosts beat the replicas, by a 9% gap. Every creator opted in and reviewed the final audio, and no listener complaints were reported.

The industry will read the 26% as proof that cloning works. I read the 9% as proof of what the study had just said: the voice people trust is the asset, and the closest copy still lost to the original. Joycast edits with human editors for the same reason: what you are selling is a person's judgment, and in this test the person won. If you read your own ads, you are already using the best-performing format in the study.

Measurement: judge a campaign at a month, not a week

Magellan AI's Q2 2026 Podcast Measurement Benchmark Report covers campaigns measured between 1 January and 30 June 2026, and the midpoint campaign put 2.23% of unique reached listeners on the advertiser's site within a 30-day lookback. The line Podnews pulled out on 11 September is the one for anyone reporting to a sponsor: 41% of site visits happen in the first week and 76% by the end of the first month, so a campaign judged after one week shows less than half of what it delivers. Tell the sponsor that before the campaign starts.

Podscribe's change follows YouTube's 24 August view change, covered in the last edition. Podscribe says view counts have risen to around double for some videos, and from Monday 21 September it will automatically report YouTube engaged views for opted-in publishers and use them as the billable metric, estimating them where a publisher has not connected. Podnews reported it on 16 September. Right call: the meaningful number stays the number you are billed on.

Then a smaller change with a sharp edge, also from Podnews on 11 September. As of 8 September, Spotify's Megaphone requires at least three published episodes before a show can be submitted to SPAN, the Spotify Audience Network, where one used to be enough. Podnews calls the change undocumented and says the company ignored requests for comment; a limited-series creator said "having to wait to monetize halfway through the run is pretty debilitating". Launching a limited series with SPAN in the plan? Budget for the ad money starting no earlier than episode three, or release the first three together.

In brief: a shortlist and automatic clips

The Independent Podcast Awards 2026 shortlist landed on 16 September, reported by PodcastingToday: 150 shows from nearly 550 entries across 28 categories, 85 judges, fourth year. Founded by Verbal Diorama and whynow, with YouTube as headline partner, the winners are revealed on 4 November at Kings Place in London, in person and by livestream.

Headliner announced YouTube Video Clips Automation on 14 September: point it at a YouTube Podcast Playlist and it clips the back catalogue and every new episode, captions and posts them automatically, currently free. My view, and only a view: automated clipping is fine for volume, but the clip that travels is usually the one a human chose, because choosing is the job. That is the whole case for short-form clips cut by someone who watched the episode.

The numbers

From the edit desk: how do you ask a guest to do something without it being awkward?

Across well over a thousand episodes, the faults I fix most often were not caused by microphones. They were caused by a host who did not want to be rude. A guest fluffs a line, the host lets it go, and there is no clean take to use. A guest joins from a noisy café because nobody asked them to move. A laptop fan whirrs through the recording and the headphones stay in the drawer. An answer runs eight minutes because the host was too polite to redirect. A guest never shares the episode, because nobody asked. Every one of those is an ask that did not happen, so move the ask to where it stops being awkward.

1. Ask in writing, before the day. The pre-call email does the awkward bit: "Quiet room, headphones in, laptop on mains and away from the mic if you can. It makes a real difference to how you sound."
2. Frame every request as being for the guest's benefit. "So you sound your best" is not a trick; it is true.
3. Get permission to interrupt once, at the top: "If I jump in or ask you to say something again, it's only so it lands better for you."
4. Make the retake ordinary. "Could we try that again, shorter?" said in a normal voice is a sentence most guests are relieved to hear.
5. Ask for the share with specifics. Not "please share it", but "it goes out on the 24th, here is the clip and a ready-made post, would you put it up that morning?"

This week's news makes the case for me. The Podcast Study says listeners come for the host, so the host's relationship with the guest is part of the product, and Apple's Search in Show means a guest's name said clearly on mic becomes a way into your catalogue. The host who asks well gets a better recording, a happier guest and a better search result, and video and audio editing is where the retakes become one clean answer.

Which conversation do you need to have before your next recording?

Probably the one with the guest, and it is a two-line email. Then say the guest's name on mic and write chapters that describe the episode, because Apple's search bar now reads both; tell any sponsor, before the campaign starts, that a week shows less than half of what it will deliver; and if you read your own ads, keep reading them. If the editing is the part you would rather hand over, a full episode edit starts at £280, billed per four-week cycle, with no minimum and revisions until you are happy. Get in touch and tell us about the show. Back next Thursday.