AI video rights checklist. Ship cleared work, every time.
Before any AI video ships, run this checklist. It covers likeness, music, source material, synthetic-media labels, client use, and the platform AI-label settings. Every platform rule links to the primary source so you can re-check it yourself.
Where these rules come from. Dated primary links.
The platform requirements here are quoted from the official help pages and announcements, fetched 29 Sep 2026. Nothing on this page is a guess; each requirement sits next to a link to the page that states it. Re-fetch the links before a client delivery, because platform policies change.
Six checks. Before the export.
01 / Likeness
Real face or voice needs written consent
If your video makes a real person appear to say or do something they didn't, that is a disclosure trigger on YouTube and a legal question everywhere else. Get written consent before you generate anyone's face or voice, including your client's staff. If no real person exists, disclose that too. Example wording for a synthetic character: "original fictional adult AI character created for this MakeRoll demo."
02 / Music
AI generation does not clear a song
If a track is in your output, you still need a license or a rights-safe source. Keep the receipt in the project folder.
03 / Source material
License the footage you prompted with
If you trained a style on, or prompted with, someone else's footage or images, check the license before the client runs it as an ad. Cite what the model saw only if you actually know.
04 / Platform labels
YouTube, TikTok and Meta each have their own setting
The three platform sections below give the exact requirement and the link.
05 / Client use
Write the clearance into the brief
Disclosure settings affect paid distribution. Put who labels, where, and who keeps the receipts into the brief before the first render.
06 / Ledger
Keep the paper trail in one folder
Consent, license, label decision, and the fetch date of each rule. If a platform changes policy, the fetch date tells you which claim to re-check first.
YouTube. Label realistic altered or synthetic content.
Trigger
When YouTube requires the label
Check the box when content looks realistic and was made with altered or synthetic media, including a real person appearing to say or do something they didn't, footage of a real event or place that didn't happen, or a realistic scene that didn't actually occur. The setting lives in YouTube Studio under the altered content question.
No label needed
What stays exempt
Beauty filters, color or lighting fixes, background blur, and vintage effects are fine without a label. So are production help like scripts, captions, and ideas. Clearly unrealistic content like animation does not need the label either.
Primary source: YouTube disclosure requirements for altered or synthetic content, support.google.com, fetched 29 Sep 2026. The same YouTube help page lists both triggers and exemptions.
TikTok. One toggle, applied at upload.
Trigger
When TikTok requires the label
Turn on the AI-generated content toggle when any part of the video was made or edited with AI, including realistic images, video, or audio. TikTok applies a "Creator labeled as AI-generated" label, and realistic AI content that also shows realistic scenes gets additional restrictions.
Edge case
Label even when the tool forgot
Some generation tools do not attach AI metadata. The responsibility stays with the uploader, so treat the toggle as a manual step in your export checklist, not something the tool does for you.
Primary source: TikTok Creator Academy, AI-generated content label, fetched 29 Sep 2026.
Meta. Disclose in the post or the caption.
Trigger
When Facebook and Instagram require disclosure
Disclose AI-generated or AI-modified video and audio in a post or caption, or through the platform's own AI-info labels. Meta applies "AI info" labels to such content, based on industry signals and self-disclosure. Labeling requirements are strictest for ads about politics, social issues, elections, and housing, employment, or credit.
Two paths
Self-disclosure vs detected labels
Meta labels can come from the creator's own disclosure or from Meta's detection. Either way the viewer sees the same AI info tag, so do not rely on Meta's detection to catch what you could disclose yourself.
Primary source: Meta Newsroom, April 2024, Meta's approach to labeling AI-generated content and manipulated media, fetched 29 Sep 2026.
Print this. Ship with it.
1. Written consent for every real face and voice, including the client's.
2. License or rights-safe source for every song, with the receipt saved.
3. License check on every third-party clip or image you prompted with.
4. YouTube: Studio altered-content question, checked when it looks realistic.
5. TikTok: AI-generated content toggle at upload, every time, not just when the tool remembers.
6. Meta: AI disclosure in the post or caption for video and audio.
7. Client brief states who labels, where, and who keeps the receipts.
8. One folder per project holds consent, licenses, label decisions, and the date you fetched each platform rule.
Built from shipping, not from theory.
MakeRoll runs weekly AI video and UGC production, and every delivery in the community ledger follows this checklist. The rules here are the boring half of quality: the camera is optional, taste is not, and neither is disclosure. If a platform changes a requirement, this page gets updated and the fetch date above tells you how fresh each rule is.
Disclosure is part of the craft. Cleared work ships faster.