Published 2026-08-22

You are about to run AI-made creative under your own brand name across Meta, TikTok, YouTube and Snap. The question is not whether that is allowed — it is what you must declare, who declares it, and what your supplier has to hand you in writing before the media plan goes live.

01Kuwait Army02Gulf Bank03Midana04Warba Bank05Dabdoob06Boursa Kuwait07Palmy08World Defense Show09Hyundai10Kellyville

In brief

• YouTube requires creators to disclose when AI meaningfully alters or generates photorealistic content, and exempts non-realistic AI content and minor edits to realistic content.
• Meta requires political and social-issue advertisers to disclose photorealistic or realistic-sounding content created or edited with third-party generative AI tools. Its stated behaviour is rejection for non-disclosure.
• TikTok allows significantly edited media and AI-generated content in ads when the AIGC label, or a clear disclaimer, caption, watermark or sticker, is applied.
• A valid C2PA manifest proves a file carries signed, untampered statements about how it was made; it does not prove the ad is accurate or compliant.
• No GCC regulator page setting out AI-disclosure requirements for advertising was verified in this pass; the UAE's official media-regulation page does not address AI-generated content.

One test decides most of it: is the content realistic?

Read the four policy sets side by side and the same hinge appears in all of them. Not was AI used. Not how much AI was used. The question is whether a viewer could mistake the frame for a recording of something that actually happened.

YouTube states the requirement plainly: "we require creators to disclose when they use AI to meaningfully alter or generate photorealistic content." It adds that "Creators don't need to disclose non-realistic content that's made with AI, or edits to realistic content that are minor." A stylised 3D loop of your product is a different compliance object from a synthetic woman in a Salmiya kitchen saying she bought it.

That distinction should already be sitting inside your approval workflow, not discovered at upload.

Realistic content is content a viewer could mistake for a recording of something that actually happened. Read the four policy sets side by side and the same hinge appears in all of them.

What each platform's own page says today

Meta. There is no blanket advertiser disclosure duty in the pages we could read. There are two narrower, hard ones. One covers political and social-issue ads. Advertisers "must also disclose when a social issue, elections, or political ad contains a photorealistic image or video, or realistic sounding audio, that was created or edited using third-party generative AI tools." Meta adds: "If we determine that an advertiser has not disclosed as required, we will reject the ad, and repeated failure to disclose may result in penalties against the advertiser." For organic posts, Meta's misinformation standard requires people to disclose photorealistic video or realistic-sounding audio that was digitally created or altered. The disclosure runs through its AI-disclosure tool. Separately, Meta labels rather than asks. It applies an AI label when its own generative features "result in a significant edit to the image or video." It has also committed to "automatically detecting ads created or edited using third-party AI tools through industry-standard signals." The label reads "AI info" and sits in About this ad. It moves next to the Sponsored label when a photorealistic human is involved.

TikTok. "We require people to label realistic AI-generated content on TikTok." On the ads side, the policy is permissive but conditional. "Significantly edited media and AIGC content is allowed if the following requirements are met: Apply the AIGC label, or by adding a clear disclaimer, caption, watermark, or sticker." TikTok defines the category as images, video or audio that are completely AI-generated or real source material significantly modified by AI.

YouTube and Google. Beyond the creator duty above, Google's election-ads policy is the strictest instrument we found. "Advertisers must disclose all election ads that contain synthetic or digitally altered content by selecting the checkbox in the 'Altered or synthetic content' section in their campaign settings." Resizing, cropping, colour correction and red-eye removal are named as out of scope.

Snap. On the pages we read, Snap's ad policies do not impose a general AI label. They ban outcomes. Section 4.3 prohibits "manipulating content for false or misleading purposes (whether through generative AI or through deceptive editing)." Section 2.4 bars using a person's "name, likeness (including look-alikes), voice (including sound-alikes)" without consent. Section 3.10 bars synthetic avatars and manipulated likenesses in political advertising.

AIGC is TikTok's category for images, video or audio that are completely AI-generated, or real source material significantly modified by AI. In ads TikTok allows it if the AIGC label, or a clear disclaimer, caption, watermark or sticker, is applied.

Labelling a synthetic presenter and fabricating a testimonial are different failures

Brands collapse these two, and the collapse is expensive. A synthetic presenter is a transparency question: the platform wants the audience to know the face is generated. A fabricated testimonial is a truth question, and no label fixes it.

Google's unreliable-claims policy is explicit: "Testimonials that claim specific results must include a visible disclaimer stating that there is no guarantee of specific results and that the results can vary." Its misrepresentation policy states that "Manipulating media to deceive, defraud, or mislead others is not allowed." Its misleading-representation rules bar "Implying affiliation with or endorsement by another organization, brand, or private citizen without their knowledge or consent."

An AI label proves the pixels were generated. It does not prove the sentence spoken over them is true, substantiated, or yours to say.

A fabricated testimonial is a testimonial-shaped line describing an experience that never happened. It is a truth question rather than a transparency one, and no label fixes it.

A generated spokesperson is not a customer

The most common defect we see in AI-made UGC-style ads is not a rendering artefact. It is a script. A generated person says "I ordered this last week and it changed my routine." Nobody ordered anything. There is no last week.

Disclosing that the presenter is AI does not convert an invented experience into a permissible claim — it just tells the viewer the liar is synthetic. If a line would require a real customer, a receipt or a study to be defensible, it still requires one. Write generated characters as presenters, hosts or brand-owned characters, and keep the claims where you can substantiate them. That is a rule for any virtual character programme, whatever platform carries it.

An Arabic in-ad disclosure that reads naturally rather than as a translated string: «يحتوي هذا الإعلان على محتوى مُنشأ بالذكاء الاصطناعي». Write it, do not translate it — the same discipline as any bilingual workflow.

Content Credentials: what C2PA proves, and what it does not

C2PA is the provenance standard the platforms increasingly read. It "provides an open technical standard for publishers, creators and consumers to establish the origin and edits of digital content." It describes Content Credentials as functioning "like a nutrition label for digital content." Version 2.3 of the specification is dated December 2025. It defines a C2PA Manifest as the set of information about the provenance of an asset built from assertions, a single claim, and a claim signature.

The specification is unusually honest about its own limits. C2PA specifications "SHOULD NOT provide value judgments about whether a given set of provenance data is 'good' or 'bad,' merely whether the assertions included within can be validated as associated with the underlying asset, correctly formed, and free from tampering." Trust rests on "the identity of the signer."

So a valid manifest proves the file carries signed, untampered statements about how it was made. It does not prove the ad is accurate, compliant, cleared for the likeness it shows, or approved by your legal team.

TikTok reads these signals. It describes C2PA as technology that "embeds metadata into content that lets us—as well as other platforms who use C2PA—know when something is AI-generated." TikTok says C2PA has helped label over 1.3 billion videos to date. Meta says it labels when it detects industry standard AI image indicators or when people disclose at upload.

A C2PA Manifest is the set of information about the provenance of an asset, built from assertions, a single claim, and a claim signature. A valid manifest proves the file carries signed, untampered statements about how it was made.

What to require from a supplier in writing

Capability decks do not survive an ad-review queue. Contract language does. Five clauses, usable against any studio — ask them of THE UNKNOWN BRAND too, and check the answers against published work and the evidence register.

• Generation inventory. A written statement, per deliverable, of which elements are fully generated, which are AI-edited real footage, and which are captured. This is the input to every platform toggle.
• Provenance handling. Whether Content Credentials are attached, preserved or stripped in the supplier's export and post pipeline, and who owns the signing identity.
• Likeness and consent. Confirmation that no real person's face, voice, look-alike or sound-alike appears without documented consent — Snap's policy names look-alikes and sound-alikes specifically.
• Claim substantiation. Every spoken or on-screen claim mapped to its source. Any testimonial-shaped line mapped to a real, identified customer, or removed.
• Rework on rejection. Who pays and how fast when an ad is disapproved for an AI-related reason. Silence here means you pay.

Suppliers who cannot produce these in a day are telling you something. So is a supplier who has never had an ad disapproved — that usually means low volume, not high compliance. The same posture applies when you select the agency in the first place.

A generation inventory is a written statement, per deliverable, of which elements are fully generated, which are AI-edited real footage, and which are captured. This is the input to every platform toggle.

What actually happens at ad review

Two systems run, and brands confuse them. Detection is automatic and applies a label. Enforcement is a policy decision and stops the ad.

Meta's stated behaviour in the political lane is rejection for non-disclosure, with penalties for repeat failure. Google's unreliable-claims policy takes a softer line on account-level action: "Violations of this policy will not lead to immediate account suspension without prior warning. A warning will be issued at least 7 days prior to any suspension of your account." Its misrepresentation policy is harsher, treating deceptive manipulation as grounds for suspension.

Plan for it operationally. Disclose at campaign setup rather than after a disapproval. Keep the generation inventory attached to each asset, so an appeal can be answered in an hour. Do not launch a full flight on a single unreviewed creative. Your media team should treat an AI-related disapproval as a documentation problem with a known answer, not an emergency. Where a claim genuinely needs a real customer, a real location or a real product in real light, that is the point to consider a physical shoot.

What we could not verify

We did not find a GCC regulator page setting out AI-disclosure requirements for advertising that we could read and quote directly in this pass. The UAE's official media-regulation page sets content standards and penalties without addressing AI-generated content. Treat any secondhand summary of a Kuwaiti, Saudi or Emirati AI-advertising rule as unverified until you have the authority's own page in front of you. National advertising and permit regimes apply to your campaign regardless of what any platform allows.

Platform policies also change without notice. Everything quoted here was read from the platforms' own pages on the date of publication. Re-read them before a major flight, and make your supplier prove they did too. Our next decision on this desk is which of your live campaigns has an undisclosed synthetic presenter in it right now.

Sources

Every source below was opened and its quoted wording checked against the live page. A source we could not re-fetch was dropped, not softened. Where a rule could not be confirmed from the body that issues it, this guide says so rather than describe it from memory. Editorial responsibility sits with the studio, not an individual author. Found a moved link or a wrong citation? Email hello@theunknownbrand.com with the URL. We will correct the page and its modification date.

1. YouTube Help (Google) — YouTube’s creator disclosure duty for realistic AI content, the exemptions, and the consequences of not disclosing
2. Meta Transparency Center — Meta’s mandatory AI disclosure for social issue, elections and political ads, and rejection/penalties for failure to disclose
3. Meta Transparency Center — Meta’s organic-post AI disclosure duty and its label-vs-remove distinction for manipulated media
4. Meta Newsroom — When Meta applies an AI label to ads, where the label appears, and its commitment to detect third-party AI tools via industry-standard signals
5. Meta Help Center — What counts as a significant generative edit in ads and where the AI info label sits
6. Meta Transparency Center — The two triggers for Meta’s AI info label: detected industry-standard indicators and user self-disclosure
7. Meta Newsroom — The label’s current name
8. TikTok Newsroom — TikTok’s labelling requirement, its use of C2PA Content Credentials, and the volume of content labelled
9. TikTok Newsroom — The origin and scope of TikTok’s AIGC labelling requirement
10. TikTok Advertising Policies — TikTok’s conditional permission for AIGC in ads and its likeness rule
11. TikTok Ads Manager Help — TikTok’s definition of AI-generated, synthetic or manipulated media for advertisers and the disclaimer mechanism
12. Google Advertising Policies Help — Google’s mandatory altered-or-synthetic-content disclosure for election ads and the exempt edits
13. Google Advertising Policies Help — Google’s testimonial disclaimer requirement and its warning period before suspension
14. Google Advertising Policies Help — Google’s prohibition on manipulated media used deceptively
15. Google Advertising Policies Help — Google’s rule against implying endorsement or affiliation without consent
16. Snap Inc. Advertising Policies — Snap’s likeness-consent rule, its ban on synthetic avatars in political ads, and its deceptive-content rule covering generative AI
17. Snap Inc. Community Guidelines — Snap’s expectation of responsible AI use and enforcement posture on organic content
18. C2PA (Coalition for Content Provenance and Authenticity) — The C2PA manifest definition, the specification version and date, and the explicit limit on what provenance validation establishes
19. C2PA — How C2PA describes itself and Content Credentials
20. The Official Platform of the UAE Government (u.ae) — The honest negative finding that this official UAE media-regulation page sets content standards and penalties without addressing AI-generated content

Frequently asked questions

متى تضع علامة «مولَّد بالذكاء الاصطناعي»؟ ما تطلبه كل منصة بنصّها — هذا الدليل بالعربية.