Hello, this is Attorney Eom Sang-Yoon of Cheongchul Law Firm.

AI-generated 'virtual influencers' such as Shinhan Life's 'Rozy,' SK Telecom's 'Nasua,' and LG Electronics' 'Kim Reah' are increasingly used as advertising models, and their content is rapidly spreading across SNS, blogs, and video platforms. However, because virtual humans depict non-existent personas as if they were real people, there is an inherent risk that consumers may mistake the endorser shown in the ad for an actual person.

With this concern in mind, on May 26, 2026, the Korea Fair Trade Commission ('KFTC') prepared a revision to the 'Review Guidelines on Endorsement and Recommendation Labeling and Advertising' (the 'Guidelines'), which took effect on June 1, 2026. The core of the revision is to explicitly designate AI-generated virtual humans as a new type of endorsement entity, and to provide concrete disclosure language and methods that advertisers must follow when using them. Today, we examine the main contents of the revised Guidelines and the practical implications that advertisers should heed.

[Key Contents of the Revised Guidelines]

Previously, the Guidelines set out unfair-conduct standards based on four endorsement entities (consumers, celebrities, experts, and organizations/institutions), with the framework centered on the obligation to disclose economic relationships between advertisers and endorsers (so-called 'undisclosed-ad' regulation). However, there was no separate rule governing situations in which the endorser itself is an AI-generated virtual human, and critics noted there was insufficient basis to find unfairness even when an advertiser did not disclose the virtual nature of the endorser.

The revised Guidelines add the concept of artificial intelligence under Article 2(1) of the Framework Act on the Development of Artificial Intelligence and Establishment of a Trust Foundation to 'III. Definitions,' and newly establish '5. Endorsement and Recommendation by Virtual Humans' under 'V. Detailed Review Guidelines.' The main contents are as follows.

1. Disclosure Obligation for Virtual Humans

Where a virtual human ('virtual human') that is difficult to distinguish from a real person is generated based on AI and used for endorsement or recommendation, the advertiser must disclose that the endorser is a virtual human generated by AI. This is a key provision that blocks the structural risk of consumer misperception inherent in virtual-human ads, and effectively imposes a new labeling obligation for virtual-human advertising.

2. Clear and Easy Recognizability

When disclosing that an endorser is a virtual human, the disclosure must be made in a manner that consumers can clearly and easily recognize. The Guidelines provide concrete labeling methods by media type, as illustrated below.

1. Text-based media such as blogs and internet cafes: Display phrases such as 'This post contains a virtual human generated based on AI' or 'Contains virtual human' in the title or at the beginning of the post. Especially for titles, it is recommended to place the marker at the very front of the title in the form of [Contains Virtual Human] so that the disclosure is not truncated in mobile environments.

2. Visual media such as photos and videos: Display markers such as [Virtual Human] in close proximity to the virtual human while the virtual human appears, using colors that contrast with the background to secure visibility.

3. Obligation of Consistency with Experiential Facts

If a virtual human endorses a specific product as if based on actual use or experience, but the content does not actually correspond to such experiential facts, this may constitute unfair labeling/advertising. Since a virtual human is inherently incapable of having actual user experience, it is not enough for the advertiser to merely add a 'contains virtual human' disclosure; the truthfulness of the endorsement content itself must also be ensured—this is the purpose of the revised Guidelines.

The existing 'undisclosed-ad' regulation requiring disclosure of economic relationships between advertisers and endorsers remains unchanged. Note that for ads using virtual humans, (i) the virtual-human labeling obligation and (ii) the economic-relationship disclosure obligation apply concurrently.

[Practical Implications]

The enforcement of the revised Guidelines is expected to bring significant changes to advertising practices using AI virtual humans.

1. First, advertisers using AI virtual humans as ad models must establish internal compliance systems to consistently fulfill the virtual-human labeling obligation across all media (company websites, official SNS accounts, influencer collaboration posts, YouTube/Instagram videos, etc.). In photo and video media in particular, the disclosure must be maintained throughout the appearance of the virtual human, so the disclosure phrase must be reliably embedded in the video from the production stage. Furthermore, if a virtual human makes statements premised on experiential facts—such as 'I used it and liked it' or 'I was satisfied after trying it'—the advertiser must separately verify whether such content actually corresponds to objective facts such as demonstrations or tests conducted by the advertiser.

2. Second, influencers and content creators who are commissioned by advertisers to produce AI virtual-human ads must be aware that the same labeling obligation applies when they post the ads on their own channels. In addition to existing economic-relationship disclosures (e.g., 'sponsored,' 'paid promotion'), the virtual-human disclosure must also be added; the labeling obligations have become multi-layered compared to before.

3. Platform operators may also bear a degree of monitoring responsibility for AI virtual-human ads distributed within their services. The KFTC stated in its press release that it 'plans to conduct monitoring to encourage correction of labeling/advertising that does not comply with the Guidelines,' so it is advisable for platforms to prepare voluntary guidelines and post-hoc correction systems for posts missing the disclosure.

Although the Guidelines are administrative rules without direct legal binding force in themselves, sanctions for violations ultimately depend on the KFTC's individual judgment as to whether conduct constitutes unfair labeling/advertising under Article 3(1) of the Act on Fair Labeling and Advertising. Therefore, mere formal non-compliance with the labeling methods in the Guidelines does not automatically result in a finding of unfair labeling/advertising, but failure to follow the Guidelines' methods substantially increases the likelihood that the advertiser will be evaluated unfavorably in the unfairness determination.

4. Finally, because the revised Guidelines apply immediately upon enforcement without a separate grace period, advertisers should promptly consider supplementary labeling measures even for existing advertising materials produced or posted before the effective date. How to strike a balance between the industrial potential of AI virtual-human advertising and the public-interest demand for preventing consumer confusion will gradually be clarified through future operational examples and the KFTC's enforcement trends.