What does the law actually require?
Article 50 of Regulation (EU) 2024/1689 covers four situations. For a clinic or service business, two matter most:
- AI systems that talk directly to people — a website chatbot, a WhatsApp assistant, a voice agent on the phone — must make it clear the person is interacting with AI, unless that is already obvious.
- Deepfake content — image, audio or video generated or altered to look real — must be disclosed as artificially generated.
The obligations apply from 2 August 2026, and the European Commission adopted interpretive guidelines on 20 July 2026. Formally, some duties sit with the technology provider. In practice, it is the clinic that puts the system in front of the patient — and its reputation is on the line.
What does a disclosure look like that doesn’t kill the booking?
You don’t need a legal paragraph. You need one good sentence:
Hello, I’m the clinic’s virtual assistant. I can book you in or answer questions about treatments. If you’d rather speak to a colleague, just say “colleague” at any time.
Three elements: say what it is, say what it can do, offer a way out to a human. People relax when they know they are in control — and a relaxed patient books more easily.
What we never leave to AI in a clinic
- Diagnosis or treatment recommendations for a specific case.
- Promises of results or final prices without a consultation.
- Collecting health data without a clear legal basis — health data is a special category under GDPR.
- Tense conversations: complaints, emergencies, unhappy patients. A human answers those, immediately.
Why it’s an advantage, not a burden
Missed calls remain one of the most expensive leaks in a service business. A well-built AI agent answers at 10 p.m., on a Sunday or during a procedure. But it only works if the patient trusts it, and transparency builds exactly that.
Clinics that hide their use of AI will be found out sooner or later. Those that say openly “we use AI to answer you immediately, and a person takes over when needed” turn a legal duty into a quality argument.
Quick checklist
- 1List every point where a customer can talk to AI: website, WhatsApp, Messenger, phone, email.
- 2Add the disclosure to the first message or the first seconds of the call.
- 3Set up the handover to a human and test it every week.
- 4Write the list of topics AI must not handle and review transcripts monthly.
- 5Ask your provider for documentation on how it marks generated content, and keep it.