Region · AI content creation

AI content creation in South Korea.

South Korea regulates AI output directly. The AI Basic Act, in force since 22 January 2026, makes AI providers label generated content and clearly mark realistic deepfakes, although advertisers who only use AI tools carry no duty under it. Advertising rules go further for brands: since June 2026 the Fair Trade Commission has required AI-generated virtual endorsers to be labelled as virtual persons, on top of strict disclosure of paid endorsements.

The AI Basic Act: who has to label what

The Framework Act on the Development of Artificial Intelligence and Establishment of Trust took effect with its Enforcement Decree on 22 January 2026. Article 31 requires AI business operators to tell users in advance when a service uses generative AI, to indicate that outputs are AI-generated, and to label deepfakes clearly, with softer treatment for artistic work. Ordinary outputs may carry invisible watermarks or metadata plus a notice on export; deepfakes need a visible label. Official guidance treats people who only use AI output in their own content as users, not operators, so a brand making ads with AI tools has no Article 31 duty. A brand that offers an AI feature to customers, such as a chatbot or virtual try-on, becomes an operator.

Endorsements, the 'virtual person' label and publicity

The Korea Fair Trade Commission's guidelines on endorsements, tightened after the 2020 scandal over undisclosed 'backdoor' advertising, require paid relationships to be disclosed clearly in Korean, and since December 2024 blog and other text posts must carry the disclosure in the title or opening. Since 1 June 2026, endorsers generated by AI must be labelled 가상인물, 'virtual person'. The rules sit under the Act on Fair Labeling and Advertising, and the advertiser is responsible. Since June 2022, the Unfair Competition Prevention Act has also protected well-known people's names, likenesses and voices against unauthorised commercial use, so an AI lookalike of an idol or actor is a legal problem as well as a fan relations one. Health functional food ads must pass self-regulatory pre-review before they run.

Hangul, speech levels and the K-content bar

Hangul groups letters into syllable blocks, and image models still produce convincing-looking nonsense when asked to write it, so generate imagery clean and set Korean as live type in a font with full coverage. Korean grammar encodes politeness. Most consumer advertising uses the polite 해요 style, formal 합니다 endings suit finance and official notices, and casual speech is reserved for brands that have earned that closeness; switching levels within an asset reads as careless. Audiences are used to highly produced visuals from K-pop, drama and K-beauty, and they spot generated artefacts such as warped hands, garbled signage or plastic skin quickly. Invest the effort in finish, not in volume alone.

Naver, Kakao, Coupang and personal data

Search and research run through Naver, whose blogs, cafés and Smart Store listings shape purchase decisions, alongside YouTube and Instagram. KakaoTalk is on nearly every phone, and brands use Kakao channels for messaging, coupons and gifting. Coupang leads online retail with its own detailed listing standards. Plan formats for these ecosystems rather than exporting a Western social set. The Personal Information Protection Act, overseen by the Personal Information Protection Commission, requires separate consent for sensitive information such as biometric data. Synthetic White's saved characters keep a virtual endorser identical across every asset, and the approval step is the natural place to confirm the 가상인물 label is present before anything is published.

Updated 25 September 2026 · General information, not legal advice. Rules change, so check the current text with your legal team before relying on it.

Questions

AI content creation in South Korea, answered.

Does the AI Basic Act require brands to label AI-generated ads?

Not directly. Its labelling duties fall on AI business operators, and official guidance treats those who only use AI output in their own content as users. The tools you use must label their outputs, however, and realistic deepfakes need visible labels. Separately, advertising rules now require AI-generated endorsers to be labelled as virtual persons.

How must paid endorsements be disclosed in Korea?

Clearly, in Korean, and for blog and other text posts, in the title or opening since December 2024 rather than among hashtags at the end. Since June 2026, endorsers generated by AI must also be labelled as virtual persons. The advertiser is responsible for disclosure, whether the endorser is a human creator or a synthetic one.

Can we create an AI idol or K-pop-style character for a campaign?

An original character, yes, and it can become a durable brand asset. A character that resembles a real idol or actor risks claims under the publicity provisions of the Unfair Competition Prevention Act and portrait rights. If the character endorses products, label it as a virtual person.

Get started

See your AI studio generating this week.

Thirty minutes. A live studio in your colours, your brand hub loaded, a real campaign brief.

See plans