According to the European Commission (checked 19 October 2026), Article 50 of the EU AI Act applies from 2 August 2026: realistic AI-made images, video and audio are to be disclosed, and AI chats are to say they are AI. A good label is small, consistent and impossible to miss. Here is what that means on a thumbnail, in a player and in a chat.
Article 50 of the EU AI Act sets transparency duties that, according to the Commission’s FAQ (checked 19 October 2026), apply from 2 August 2026. Under the AI Act (Art. 50(4)), if you publish AI-generated or manipulated images, video or audio that look real, you are expected to disclose that they are AI-made. For evidently artistic, creative or fictional work, the Act describes a lighter disclosure, one that doesn’t spoil the work.
Article 50(1) and (2) add two related duties: people must be told they are talking to an AI system, unless it is obvious; and providers of AI systems that generate content must mark outputs in a machine-readable way. According to the Commission, providers of generative systems placed on the market before 2 August 2026 have until 2 December 2026 for that marking. Article 50(2) addresses providers, so it is relevant where your own generator makes the files.
AI content is now common on adult sites: generated scenes, AI-edited bodies and faces, synthetic voices and AI companions. Visitors often can’t tell. A clear label helps visitors tell AI from real, and helps keep real performers from being confused with generated content.
The legal line around sexual deepfakes is also moving. According to legislation.gov.uk (checked 19 October 2026), creating a purported intimate image of an adult without consent has been an offence in England and Wales since 6 February 2026, and the US TAKE IT DOWN Act covers digital forgeries. A label doesn’t make such content legal; on legitimate AI content it tells people what they are looking at.
Show a notice at the start of AI videos: a first frame or a short overlay saying the video is AI-generated. Keep the badge visible in the player chrome while it plays.
If you generate the files yourself, switch on machine-readable marking in your pipeline: C2PA content credentials, or the IPTC field DigitalSourceType set to “trainedAlgorithmicMedia”. That marking travels with the file when it is downloaded or reposted.
The first message should say it plainly: “Hi, I’m Mia, an AI character, not a real person.” Next to the character’s name, keep a fixed label “AI character” for the whole conversation. Say it on the chat’s landing page too, before anyone pays or signs up.
Avoid wording that suggests a real performer is on the other end. If real people sometimes take over a chat, say when that happens.
Do AI-edited images count, or only fully generated ones? The rule covers generated and manipulated content that looks real. If AI changed a face, a body or a voice in a way a viewer would not expect, label it.
Is “AI” in the category name enough? It helps, but people arrive from search, links and shares straight onto single pages. The label should sit on each item, not only on the category.
What about illustrated or animated content? For evidently artistic or fictional work, Art. 50(4) describes a lighter duty: a disclosure that does not spoil the work. A small consistent badge is one way to give that lighter disclosure.
Do I need a label on AI chat replies themselves? The AI Act (Art. 50(1)) says people must know they are talking to an AI. A clear first message and a fixed label next to the name usually do that better than tagging every single reply.
The easiest way to label consistently is to ask at upload time. Add a required question “Is this AI-generated or AI-edited?” to your upload form, store the answer with the item and let your templates show the badge automatically. Then a label never depends on someone remembering it.
We read your public pages by keyword for signs of AI content and AI chats, and look for visible labels and chat notices. We can’t tell whether a specific image or video is AI-made, and we never download your media, so the machine-readable part is something you tell us. We also flag words that suggest an “undress” or nudify feature, which no label can fix.
Not legal advice. This page summarises and links to primary sources; it does not tell you what the law requires of your specific site. Related: AI labels in the EU · Guide: AI labels on adult sites · Remove nudify features
One free scan shows your age check, report routes and labels, with evidence and fixes.