This page has affiliate links. Here is what that means.
The short answer, per platform
Three of the platforms you are likely buying on have landed in different places. One makes it your job, one makes it mostly theirs, and one has just started letting you put the label inside the creative itself.
| TikTok | Meta | ||
|---|---|---|---|
| Who labels a normal ad | You, before it runs | The platform, on detection | You or the platform |
| Your own disclosure required | Yes, for AIGC and significantly edited media | For political and social-issue ads | For election ads |
| Label inside the creative allowed | Yes: caption, watermark or sticker | Not the described route | Yes, since July 2026 |
TikTok: the label is yours to apply
TikTok’s ads policy allows significantly edited media and AI-generated content on one condition: you apply the AIGC label, or you add a clear disclaimer, caption, watermark or sticker of your own. Either satisfies it. There is no version of this where you do nothing.
In Ads Manager the AI-generated content disclaimer is one of three disclaimer toggles, alongside the standard and clickable kinds, and it is listed as mandatory for AI-generated, synthetic or manipulated media. That covers images, video and audio that are fully AI-generated, and real source material significantly modified by AI.
The organic side of the house works the same way for creators, and unlabelled content that needed a label may be removed, restricted, or labelled by TikTok rather than by you.
Meta: mostly the platform’s job now
Meta has moved in the other direction. When an advertiser uses Meta’s own generative AI creative features and the result is a significant edit to the image or video, Meta applies an AI label itself, either behind the three-dot menu or next to the Sponsored label.
Since mid-2026 it has also started adding information about ads it detects were made with third-party AI tools, surfaced through an “About this ad” view. So an ad built in HeyGen or Synthesia may end up labelled whether or not you say anything.
The exception is political. Advertisers must still self-disclose when AI is used to create or alter ads about social issues, elections or politics.
Google: labels you can now put in the creative
Google’s change is narrower but useful. Since July 2026 you may add text or visual AI labels directly inside image and video ad creatives, and those labels are explicitly exempt from the policies that otherwise forbid text overlays and watermarks. Google Ads may also label certain assets its own AI tools generated.
Election advertisers have a separate, older obligation: tick the altered-or-synthetic-content box in campaign settings for any election ad containing synthetic or digitally altered content.
What counts as significant
Google draws the clearest line. Resizing, cropping, colour and brightness correction, defect removal such as red-eye, and background edits that do not depict real events are all inconsequential and out of scope. The test is not whether a tool touched the file. It is whether the edit changes what a viewer would believe actually happened.
Before the ad goes live
Three checks, in order. Does the plan you are on grant commercial rights to the output? Does the creative show a realistic person or scene that AI generated or substantially changed? Have you applied the platform’s label, or written your own disclaimer into the creative?
Regional rules sit on top of all of this: Google notes that AI regulations in the European Union, India and New York already require disclosures on certain AI-generated or edited ad assets. If you run in those markets, treat labelling as the default rather than the exception.
Pickups
Does a label hurt performance?
That is an empirical question for your own account, not a policy one. The policy question is settled: an unlabelled ad that needed a label is a removal risk, and on TikTok the platform may label or restrict it for you.
Does an AI voiceover over real footage count?
Audio is in scope. TikTok's mandatory disclaimer covers images, video and audio that are completely AI-generated, and real source material significantly modified by AI. A synthetic voice reading over your own footage is the common case people miss.
What about colour grading and cropping?
Out of scope. Google names resizing, cropping, colour and brightness correction, defect removal and background edits that do not depict real events as inconsequential. The line is whether the edit changes what a viewer would believe happened.
Do I need to disclose if I used AI only for the script?
The rules are written about images, video and audio, not the writing behind them. A script drafted with AI and then performed or filmed normally is not what these policies describe.
Which disclosure is always mine to make?
Political and social-issue advertising. Meta requires advertisers to self-disclose AI used to create or alter those ads, and Google requires election advertisers to tick the altered-or-synthetic-content box in campaign settings.
Source log
Read Nov 15, 2026. These pages change without notice; the source page wins over this one.
