AI-Generated Model Ads: Disclosure Rules and How to Use Them Well (2026)
ImageFactory Engineering · Published 2026-06-18
Yes, you can run AI-generated virtual-model creative on Meta. The key is to disclose that the person is AI-generated so consumers can tell, and to avoid staging it as a real first-hand testimonial. In Korea, labeling AI virtual persons becomes mandatory on June 1, 2026 under revised Fair Trade Commission guidelines, and Meta applies its own "AI Info" label. Rules keep evolving, so check the current policy when you run.
Is labeling AI virtual persons required in Korea?
It's about to be. Korea's Fair Trade Commission revised its review guidelines on endorsements and advertising so that, effective June 1, 2026, ads featuring an AI-generated person must clearly disclose that the person is virtual (MBC report).
The method differs by medium:
- Text-first media (blogs, forums): a phrase like "virtual person included" (가상 인물 포함) in the title or opening section.
- Video / image media: a label like "virtual person" (가상 인물) near the person while they appear on screen.
Removing or tampering with the disclosure is prohibited. Confirm the exact wording and scope against the official guidance at the time you publish.
If I add a label, is any wording fine?
No — and this is where marketers most often slip. The Korean guidance notes that even when an ad discloses the person is virtual, describing the product as if that person personally used it can still be deemed deceptive advertising (MBC report).
So the "label" and the "claim" are two separate obligations. Showcasing a product beautifully with an AI model is fine; putting a first-person testimonial — "I used this for a month and it worked great" — in a synthetic person's mouth is risky. Tie real usage results and efficacy claims to genuine users or verified evidence.
How does Meta (Facebook/Instagram) handle AI disclosure?
Meta applies an "AI Info" label when industry-standard AI indicators are detected in the content or when a user self-discloses that it's AI-generated (Meta Transparency Center). Using an AI model in an ordinary product ad isn't banned in itself, but political and social-issue ads carry stricter, separate disclosure obligations when digitally created or altered, and deceptive deepfakes are disallowed outright (Meta Newsroom).
Practical tip: whether or not automated detection fires, if the content could mislead, self-disclosing reduces the risk of reach penalties. Meta's policy updates over time, so review its policy pages before you publish.
What about other markets, like the US?
The US has no single AI-specific ad law, but the Federal Trade Commission (FTC) applies existing consumer-protection law (FTC Act Section 5). AI "influencers" or endorsements that read as human can violate endorsement rules if they mislead, and the guiding principle is that disclosures must be "clear and conspicuous" (FTC AI page). Some states are introducing their own statutes, so for global campaigns, check the rules for each target country or state.
This article is general guidance, not legal advice. Confirm applicability and exact disclosure requirements against the current platform and regional rules at the time you run.
Checklist
- If a virtual person appears, did you add a "virtual person" / "AI-generated" label (mandatory in Korea from June 2026)?
- Is the label placed up front in text media and near the person in video/image?
- Does the AI model avoid first-person testimonials or efficacy endorsements (deceptive even when labeled)?
- For Meta ads, did you consider the "AI Info" label / self-disclosure? Stricter review for political/social-issue ads.
- Are factual claims (efficacy, comparisons, price) tied to real evidence?
- For international targets, did you check country/state-specific rules?
How ImageFactory helps with AI model creative
ImageFactory automates ad-creative production and adaptation. Starting from a single master or product cutout, it generates multiple style variations — including a Model style where AI picks a fitting human model — and adapts them across 1,400+ placements and 110+ platforms while preserving text and logos. It handles safe zones automatically, keeps distortion at 0-2%, and supports 2K/4K, 15 languages, and Figma/Photoshop plugins.
In other words, building virtual-model creative and rolling it out across Meta, Instagram, and many other formats is exactly what ImageFactory speeds up. But the marketer remains responsible for disclosure — always check the checklist above and the current platform/region rules. See turning a product photo into ad visuals and the supported sizes, and try it free for 14 days.