If you post AI-generated content, such as images, video or audio, to an international audience, you are now subject to AI labelling rules in several of the world’s largest markets. The details differ, but they share a core idea: people should be able to tell when content was made with AI. One well-written disclosure, plus the platform’s own AI label, covers most cases.
AI labelling rules at a glance
- China: since 1 September 2025, AI-generated text, images, audio and video must carry visible labels and hidden metadata labels. Platforms must check for them. Tracker entry.
- South Korea: since 22 January 2026, the AI Basic Act has required generative AI output to be marked and realistic deepfakes to be clearly labelled. Tracker entry.
- India: since 20 February 2026, synthetic content on platforms must carry labels and permanent provenance markers, and large platforms must ask users to declare AI content. Tracker entry.
- European Union: since 2 August 2026, deepfakes must be disclosed and AI-generated content must be marked in a machine-readable way. Generative AI tools already on the market before that date must add the marking by 2 December 2026. Tracker entry.
- California: since 2 August 2026, generative AI providers must embed hidden provenance data in AI-generated media where technically feasible, with platform duties from 2027.
- Brazil: election campaigns must label AI content, and AI fakes of candidates are banned.
Who has to label AI content: you or the tool?
Most of these laws put the main duty on AI providers and platforms, not on individual users. The AI tool adds hidden markers, and the platform shows a label. But users have duties too: in India you may be asked to declare that an upload is AI-generated, in China you must not remove labels, and under Article 50 of the EU AI Act anyone using AI to publish a deepfake, other than for purely personal, non-professional purposes, must disclose it. Creators who use AI to post realistic synthetic content should assume a label is expected.
EU AI Act Article 50: the transparency rules
The EU’s labelling rules sit in Article 50 of the AI Act, which sets out its transparency obligations. Providers of an AI system that generates synthetic audio, images, video or text must mark the output in a machine-readable way, such as a watermark or metadata, so AI-generated content can be detected. Deployers who publish a deepfake, outside purely personal use, must disclose it. For an evidently artistic, satirical or fictional work, a light disclosure that does not spoil the work is fine.
Text generated by AI and published to inform people about matters of public interest must also be disclosed, unless a person has reviewed it and someone holds editorial responsibility for it. Breaches of these AI duties can lead to fines of up to €15 million or, if higher, 3% of a company’s worldwide annual turnover, under Article 99.
Platform AI label rules apply too
Even where the law is silent, platforms have their own AI disclosure requirements. YouTube has asked creators since March 2024 to disclose realistic altered or synthetic content. TikTok requires a label on realistic AI-generated content, and Meta adds “AI info” labels when it detects industry-standard markers or when users disclose. Breaking these rules can get content removed or reduce its reach.
One AI disclosure that travels
To label AI-generated content, say what was AI-made and what was yours, close to the content itself:
- “Images generated with AI; script and editing by me.”
- “Voice-over created with AI. Footage and research by me.”
- “Made with AI, then edited and fact-checked by me.”
Then switch on the platform’s AI label as well, and do not strip metadata, such as C2PA Content Credentials, from AI-generated files before uploading. That combination meets the spirit of every rule above.
A quick AI label checklist before you post
- Is it realistic? If someone could mistake it for a real photo, recording or event, label it clearly.
- Does it show a real person? Get their consent, and never create intimate or humiliating content.
- Where is your audience? If you reach viewers in China, India, the EU or South Korea, assume a label is legally expected.
- Did you keep the metadata? Do not strip the hidden markers or watermark your AI tool adds.
- Did you switch on the platform label? Use it every time, even for partly AI-made content.
The rules do not agree on everything. Some require visible labels, others only hidden ones; some cover all AI content, others only realistic fakes. Expect guidance to keep changing, and new rules to take effect, especially in the EU and India. Our labelling rules tracker shows the current status in each country.
Last checked: 5 November 2026. This explains the law in plain English; it is not legal advice. Spotted an error? Tell us, and see how we check facts.
