Samsung is running AI-generated ads on TikTok without telling anyone. The same videos posted to YouTube carry AI disclosures in their descriptions. On TikTok? Nothing. The Verge AI investigated this transparency gap and found a systemic failure that cuts across platforms, advertisers, and the very industry groups that claim to champion AI content labeling.
What makes this particularly absurd: both Samsung and TikTok are members of the Content Authenticity Initiative (CAI), a coalition pushing for industry-wide adoption of C2PA content provenance standards. They supposedly share “similar ideals regarding the labelling of AI content,” as The Verge AI puts it. Yet Samsung’s Galaxy S26 Ultra promo videos ran on TikTok feeds for weeks with zero AI disclosure, while identical content on YouTube was properly labeled.
Samsung declined to comment. TikTok pointed to its existing policies and C2PA partnership but wouldn’t explain why Samsung’s ads got a pass.
The Policy Exists, the Enforcement Doesn’t
TikTok’s own advertising policy is actually pretty clear. Advertisers can only use content “significantly” edited or generated by AI if they disclose it. That includes:
- Completely AI-generated images, video, or audio
- Showing subjects doing things they didn’t actually do
- AI voice-cloning to make subjects say things they never said
Advertisers can use TikTok’s built-in AI label or add their own disclaimer. The tools are there. The will to use them isn’t.
The Verge AI did spot one small improvement: UK car retailer Cazoo’s ads, which previously ran without disclosure despite containing obvious AI artifacts (a dentist’s drill morphing between hands, for instance), now carry an “advertiser labeled as AI-generated” tag. Whether this was voluntary or prompted by scrutiny isn’t clear.
Why This Matters More Than You Think
This isn’t just an academic debate about watermarks and metadata standards. Advertising is a regulated industry with consumer protection rules that already exist. The EU, China, and South Korea have all introduced labeling requirements for AI in promotional materials. Companies that ignore these rules face real fines.
The comparison to cosmetics advertising is sharp. Laws prevent beauty brands from slapping false lashes on models to sell mascara. TikTok influencers have learned the hard way that audiences react badly to dishonest promotion tactics. AI-generated ads operate in the same territory: they can create scenes, products, and results that don’t exist in reality.
The deeper problem is that no trusted technological solution exists for reliably identifying AI-generated content at scale. C2PA Content Credentials, SynthID, and other provenance systems all require universal adoption to work. That’s not happening. The CAI has over 200 members, yet its own members can’t consistently label their own AI content across platforms.
What This Signals for the Industry
Three takeaways for anyone working in AI, advertising, or platform policy:
- Self-regulation is failing. Voluntary industry groups like CAI aren’t producing consistent behavior even among their own members. This strengthens the case for mandatory regulation with enforcement teeth.
- Cross-platform disclosure gaps are the norm, not the exception. If Samsung labels on YouTube but not TikTok, the labeling isn’t about transparency. It’s about whatever each platform enforces (or doesn’t).
- The enforcement burden keeps falling on users. Right now, spotting AI-generated ads requires the kind of visual forensics training most consumers don’t have. That’s backwards.
The honest read here is that nobody in this chain, not the advertiser, not the platform, has strong enough incentives to prioritize transparency over engagement. Until regulators start issuing fines or platforms start rejecting unlabeled AI ads at submission, disclosure will remain optional in practice, no matter what the policies say.
The full investigation is worth reading over at The Verge AI, especially if you’re navigating AI disclosure requirements for your own campaigns.