Apple Podcasts 1.11: disclose AI voices in the audio and the metadata
Apple's podcast guidelines require prominent disclosure of synthetic voices and AI hosts, in the content and the metadata. What to write, and a record to keep.
Apple has a specific rule for AI voices. Section 1.11 of the Apple Podcasts content guidelines, read on 2026-10-05, says creators using AI to generate audio or video content, including synthetic voices, AI-generated hosts or on-screen personas, or AI-synthesised replicas of real individuals, must prominently disclose this to audiences. It says the disclosure must be in the content and in the metadata for each episode and show.
Section 1.12 adds a ban on using AI to mislead or deceptively portray real-life events, such as fabricating news stories or manipulating audio or video clips to present a false narrative.
Two places, not one
| Place | What to do |
|---|---|
| Inside the audio or video | Say it aloud or show it on screen, prominently |
| Episode metadata | Mention it in the episode description |
| Show metadata | Mention it in the show description |
| Real-world events | Do not use AI to fabricate or falsify them |
What "prominently" asks for
Apple does not define a word count, so a plain approach is safest: a spoken line in the first minute and a line in the episode notes. For example: "This episode uses an AI-generated voice for the narration." For a video podcast, add an on-screen line in the first seconds.
If you make narration with a text-to-speech job and a talking-head with an avatar video job, you are covered by the first clause for both the voice and the on-screen persona. A replica of a real person's voice needs the strongest disclosure and, in practice, that person's permission.
A simple record
Keep the record next to the source files so that if Apple asks, you can answer in one place. Sume's docs say jobs accept a metadata field that Sume stores and does not send to the provider, which is a convenient place for an episode id. Sume's docs describe a metadata field on video, image and music jobs that Sume stores on the job and does not send to the provider. I found nothing in the docs about Sume adding, keeping or removing C2PA credentials, watermarks or platform labels, so treat that as undocumented and check the delivered file.
- Episode title and date published.
- Which parts are synthetic: narration, intro, a guest voice, visuals.
- The tool, model and job id for each synthetic part.
- The exact disclosure wording in the audio and in the metadata.
Sources
Related posts
More in Use cases
- Approve the first frame before paying for motion: a $0.0094 draft gate
Review three $0.0094 draft stills, approve one, then pay $0.0835 for the final and $1.89 for the clip: $2.00 a SKU, against $5.75 for three blind clips.
- ArtStation 25 MB free vs 250 MB premium clips: bitrate budget
A 60-second ArtStation clip on a free account must average about 3.3 Mbps to fit in 25 MB. Premium's 250 MB allows about 33 Mbps. Table and a Python check.
- ArtStation free 2K video cap: conform an AI clip with trim output
ArtStation caps free video clips at 2K and shows premium 4K in full resolution. Sume trim output takes width and height from 256 to 2160, so you can conform.
- ArtStation video clips are 1 minute: trim an AI turntable to fit
ArtStation video clips run up to 1 minute as MP4 and autoplay in a loop. Trim an AI-generated turntable to 60 seconds or less with Sume, then probe size.
Written by Sume