Do AI-generated ads have to be labeled?

Direct answerOn TikTok, yes when the AI change is significant: a fully generated scene, a person doing something they did not do, or a voice clone of words they did not say. A lighting fix or a background swap is what TikTok calls insignificant. Google does not publish that same label chart. It does ban using a synthetic public figure so people think that person endorses the product. Meta’s public ad-standards introduction does not state an equivalent label. Absence of a sentence is not a promise that unlabeled synthetic ads will be approved.

This is not the music and likeness checklist

Whether you have a license for the song, the creator’s face, or a brand logo is copyrighted music, UGC, and brand logos. A disclosure sticker does not create that license. This page starts after you know whose face and whose audio you are allowed to use.

TikTok’s split

The misleading and false content policy (updated April 2026 on the page we reviewed) allows significantly edited media and AI-generated content if you apply the AIGC label or add a clear disclaimer, caption, watermark, or sticker of your own. “Significantly modified” means more than a minor tweak. TikTok’s examples:

If you are unsure, TikTok says to use caution and label. If TikTok identifies AI-generated content that was not disclosed, the ad can be rejected or restricted.

The same article’s identity section is separate from the label. TikTok does not allow content that uses a public figure’s likeness without permission, particularly in misleading contexts. The examples include a public figure making a false endorsement or being endorsed. Private individuals’ likenesses used without permission are also described as removable once TikTok is aware. A watermark that says “AI” does not answer that section.

Google: the published problem is the fake endorsement

The Google Ads policy pages reviewed for this article do not publish TikTok’s significant-versus-insignificant label table. What Google does publish, under unacceptable business practices, is public-figure impersonation. The examples include using a famous person’s name or image without permission so they look like a happy client, falsely stating that an athlete sponsors a product, and using AI-generated videos of popular YouTubers so people think they endorse a get-rich-quick scheme.

Policy: those impersonation examples are not allowed. Best practice: treat a synthetic spokesperson who is meant to be mistaken for a real person as that policy, on every platform, not only Google. The income claim itself is income and business-opportunity ads. Fix both. A truthful label on a guaranteed-earnings script still leaves the earnings claim.

Google also says to have documented permission before using someone’s name, image, or video as if they endorsed you. That sentence is on the unacceptable-business-practices page. It is the practical test for a creator whitelisting deal: the permission should match the claim in the ad.

Meta: do not invent a label rule

Meta’s Advertising Standards introduction prohibits deceptive or misleading schemes, and the standards require ads to follow community rules on intellectual property and on adult sexual content. In the public introduction reviewed for this article, there is no AIGC-label requirement comparable to TikTok’s. Policy: we will not state a Meta disclosure duty the page does not contain. Best practice: still do not imply a real person said something they did not say, and recheck the live standards, because this topic is where pages change first.

A synthetic scene that is sexually explicit still fails Meta’s adult-content rules and TikTok’s ban on sexual services, including the “AI Nudify” example on TikTok’s adult content policy. Disclosure is not a path around that ban. Dating-app permission is a different product, covered in dating ads.

A shot list you can actually use

1. Color-corrected UGC from a creator who signed a release: TikTok’s insignificant list. Keep the release. No extra AI claim is required by the examples.
2. The same creator’s face, with a cloned voice reading a script they never recorded: TikTok significant. Label it, and make sure the release covers the new words.
3. A generated person who is not based on anyone real, holding the product: TikTok says fully generated content gets a label. Do not style them as a known founder or athlete.
4. A generated celebrity telling people the course prints money: Google’s impersonation example plus an unreliable income claim. Do not run it.

Setup workflow

1. Ask whether AI changed what the person does, says, or looks like in a substantial way, or only graded the shot.
2. On TikTok, label significant AI and fully generated scenes with the AIGC label or your own clear disclaimer, caption, watermark, or sticker.
3. If the face or voice is a real public or private person, get permission. A label does not replace permission.
4. Do not use a synthetic celebrity to imply they use the product. Google’s unacceptable-business-practices examples include that pattern.
5. If the spoken claim is an income or health promise, fix the claim. A disclosure sticker does not make a guaranteed result reliable.

Frequently asked questions

Sources

• TikTok: Misleading and false content (edited media and AIGC)
• Google Ads: Unacceptable business practices
• Meta: Introduction to the Advertising Standards
• TikTok: Adult content

Synthetic-media rules are changing quickly. This page reflects the public articles as of September 21, 2026. It is not legal advice. If a platform page adds a label requirement after this date, the live page wins.