Does that site tell people they are talking to an AI?
Enter a website. We fetch the page once, look for a chat or AI interface, and look for the wording that identifies it as an AI system — the thing Article 50(1) has required since 2 August 2026. Free, no sign-up.
What this scan can and cannot see
Every other free tool on this site asks you to describe your own product. This one looks at what is actually shipped — but through a narrow window. It reads the HTML a server returns for one URL. It does not run JavaScript, cannot open a widget, cannot log in, and cannot hear a voice agent.
So the two outcomes are not symmetrical, and we report them that way. Finding disclosure wording is good evidence the notice exists. Not finding it is not evidence of a breach — it is a reason to open the widget yourself and see what a first-time visitor is told before the conversation begins.
One more distinction that matters more than people expect: a chat box answered by a human being is not an AI system, and Article 50(1) says nothing about it. The scan separates products that are AI agents by definition from platforms that could be staffed either way, and tells you which one it found.
If the notice is missing
The fix is usually one sentence shown before the first message. The disclosure wording library has copy-paste notices for chatbots, voice agents, generated content and deepfakes in 8 EU languages, with no sign-up. To see which other Article 50 duties bind the same system — content marking, deepfake labels, emotion-recognition notices — run the Article 50 readiness checker.
Shipping the sentence is the easy half. Being able to show when you shipped it is what an authority asks for. Conformly classifies each AI system, generates its duties and keeps dated evidence against each one. Run the free risk checker →
Questions
- Does “no disclosure found” mean the site is breaking the law?
- No. The scan reads one server-rendered HTML response. Most chat widgets build their interface with JavaScript after the page loads, and the notice very often lives inside the widget itself — sometimes in a cross-origin frame that no external scan can read. An unconfirmed result means exactly that: we could not confirm it. Open the widget and check what a first-time visitor is told.
- My chat is answered by real people. Does Article 50 apply?
- No. Article 50(1) covers AI systems intended to interact directly with natural persons. A live-chat box staffed by humans is not an AI system, and the transparency duty does not attach to it. That is why the scan separates products that are AI agents by definition from platforms that may be staffed either way.
- What exactly does Article 50(1) require?
- Providers must ensure that AI systems intended to interact directly with natural persons are designed and developed so that those persons are informed they are interacting with an AI system — unless that is obvious to a reasonably well-informed, observant and circumspect person. The obligation has applied since 2 August 2026.
- Do you store the sites I scan?
- No. The page is fetched once, analysed in memory and the result is returned to your browser. Nothing about the target is written to our database. If you ask us to email you the result, we store only your address and the verdict.
- Can I scan a site I do not run?
- The scan only reads a page that is already public, the same as any browser. It is built for checking sites you are responsible for — your own, or your customers' with their agreement. It fetches one page and never crawls.