YouTube's AI label for unrealistic content: the expanded description

YouTube puts a label on realistic AI content, but unrealistic AI content is disclosed in the expanded description. The realistic test and the API field.

4 min readSume
All posts

Not every AI video gets a label on the player. According to the YouTube blog post, read on 2026-10-05, labels for significant photorealistic AI content show on the video, as an overlay on Shorts and below the player on long-form, while unrealistic content is disclosed in the expanded description.

So the first question for a creator is whether a clip counts as realistic. The second is how to set the disclosure.

What counts as realistic

The YouTube Data API revision history describes the status.containsSyntheticMedia property from its October 30, 2024 entry. It covers making a real person appear to say or do something they did not, altering footage of a real event or place, and generating a realistic scene that did not occur. The Videos resource page says the property lets the channel owner disclose realistic altered or synthetic content in videos.insert or videos.update.

Two outcomes

Where YouTube shows AI disclosure, per its blog post (read 2026-10-05)
ContentWhere the disclosure shows
Realistic AI, ShortsOverlay on the video
Realistic AI, long-formBelow the player
Unrealistic AI contentExpanded description

Sorting a Sume clip

A clip from the Sume video API of a cartoon fox is unrealistic content. A generated shot of a street that looks filmed is realistic. When you disclose through the API, the property is for the realistic category, so a cartoon does not need it, although a note in your description is cheap.

If you are not sure, treat the clip as realistic if a viewer could take it for footage. The label does not by itself change recommendations or monetization eligibility, according to the same post, so disclosing a borderline clip carries little cost.

Keep a record

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.

  • For each upload, note realistic or unrealistic and why.
  • Save the Sume job id and the model.
  • Add a description line for unrealistic AI content, matching where YouTube shows it.

If you upload through the API rather than Studio, put the disclosure decision in your upload script as an explicit field, so a batch cannot silently skip it, and log the value you sent next to the Sume job id. The Videos resource page names videos.insert and videos.update as the calls where the property can be set, so a later correction is one update call per video rather than a re-upload.

Sources

Related posts

More in Use cases

All Use cases posts

Written by Sume