Recurring AI lead in a Shorts series: label every episode?

Do you have to label every episode of a series that uses an AI lead? Assume yes for realistic episodes. YouTube asks creators to disclose when they meaningfully alter or generate photorealistic content, and TikTok requires labels on content completely generated or significantly edited by AI that looks realistic. Keep one record per episode so the answer is the same every time.

YouTube is rolling out Shorts series from September 23: episodes grouped in seasons, portrait or square, up to 180 seconds each, and a 'watch series' button that opens a series feed. A recurring lead means every episode in that feed is a candidate for a label.

What the platforms say

Both statements come from the platforms' own pages, read today.

What does not need a label on YouTube

YouTube's page lists examples that do not require disclosure: fantasy scenes, beauty filters and color adjustments, voice cloning for your own overdubs, and script or thumbnail generation help. It also says disclosure does not reduce reach or affect monetization eligibility. A realistic avatar presenter is not on that list; a cartoon character is closer to the fantasy example, but the page is the authority, not this summary.

A per-episode record

Sume returns a job per render with a job id, model and artifact URL. Store those next to the episode number, the platform label you applied and the date. If a platform changes its rule, you can answer 'which episodes used the avatar' in one query.

Put the label in the upload flow rather than burning a line into the video. Burned-in text cannot be removed if a platform adds its own automatic label.

Not legal advice

Rules change and vary by country. Treat this as a workflow note: check the page for each platform before you publish, and when in doubt apply the label.

Making it routine

A recurring character makes the disclosure question permanent, so build it into the workflow. Decide the policy once, store the answer on the episode record, and apply it at publish time. Vendors change their wording, so link the current page in your internal runbook and re-read it each quarter.

• Label each episode on upload, not once per series.
• Keep the generation job id next to the published video id.
• Disclosure on YouTube does not reduce reach, per YouTube's help page.

Mixed content in one series

Some episodes may be live footage with an AI lead-in, others fully generated. Disclosure applies at the episode level, so a series can contain both. Keep a column for the answer in your episode table, and set it from the job type: anything created with an avatar or video model gets a yes by default until a person says otherwise.