Since 2 August 2026, sites in the EU that show realistic AI-made images, video or audio of people must say so, under the AI Act (Art. 50(4)). Under the AI Act, AI chat characters must tell users they’re AI. Here’s what that means in practice, what a good label looks like, and how to check your site in a minute.
Do you have unlabelled AI content? Find out in about a minute.
Free · private · we never download your images or videos
Ask yourself these four questions. If one of these is a yes, Article 50 may apply to your site.
Machine-readable watermarks (C2PA and similar) are mainly the job of the AI tool’s maker. If you run your own generator, that part may apply to you too.
Small, always visible, same place on every card.
Two seconds before playback. Works for audio too, spoken.
The very first line says it, and a fixed label stays visible.
We read your pages, sitemaps and media lists the way a visitor would. We look for the traces AI tools leave, then check whether a label sits next to that content. We never download or look at your images.
4 file names start with ChatGPT-Image. 2 of them are visible on your home page before the age check, 2 are in your media sitemap. We found no AI label next to them.
The fix: add a badge to those thumbnails and an AI notice in your footer. Copy-paste snippet for WordPress included.
Article 50 is enforced by the authorities each EU country names. Separately, since 6 February 2026, creating a sexual deepfake of an adult without consent, or asking someone else to create one, has been a criminal offence in England and Wales (SI 2026/31, commencing s. 138 of the Data (Use and Access) Act 2025); in the US, the TAKE IT DOWN Act makes knowingly publishing such images a crime. A label doesn’t make those legal. From 2 December 2026, Article 5(1)(ba) of the AI Act, added by Regulation (EU) 2026/1744 (checked 20 October 2026), prohibits AI systems that generate or manipulate realistic intimate images, video or audio of an identifiable person without that person’s explicit consent.
Probably yes, but a lighter one. Article 50 asks for a disclosure for content that resembles real people and would falsely appear authentic. Whether wholly fictional AI performers count: the official sources don’t settle this yet. A discreet, always-visible “AI-generated” badge is a simple default and costs you almost nothing.
The rules are about content shown to people in the EU. If EU visitors can see your AI content, assume Article 50 matters for you, wherever your company sits.
Usually not. Light edits like colour correction or cropping don’t change what’s shown. Once AI changes who or what appears, or creates a realistic scene that didn’t happen, treat it as a deepfake and label it.
No. We can’t tell whether a single image was made by AI. We find the traces AI tools leave, like file names, and whether a label sits next to them. If you tell us you publish AI content, we check more carefully.
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: DSA for adult sites · Glossary: deepfake · Fix guide: AI labels
One scan checks AI labels, age checks and your report page, from the country your visitors are in. You get the findings, the evidence and copy-paste fixes.
No signup for the first result · built by people who run an adult site themselves