AI transparency
How we use AI
Some of what you see on this site is produced with the help of software that writes and reads language. We think you are entitled to know precisely where that happens, what it is allowed to touch, and who answers for it. This page says so in plain terms, and it describes what runs today — not what we plan to build.
Where AI is actually used
Our conversational assistant. When you message us on WhatsApp or through the chat on this site, an AI assistant may read what you write, answer general questions about the practice, our practitioners, opening hours, fees and how to prepare for an appointment, and pass your request to the right person. It can draft a reply for our care coordinator to check and send.
Organising requests. AI helps sort incoming messages and appointment requests so that the right practitioner or the care coordinator sees them, and helps draft summaries and administrative text that a member of our team then reviews.
Language. The assistant can read and reply in French, Dutch, English and German, and can translate what you write so that a practitioner who does not speak your language can still understand your request.
Not everything that looks clever is AI. The search box on our booking page is a plain, fixed list of words and synonyms maintained by our team — type "kiné" and it also matches "physiotherapy" because someone wrote that down, not because a model decided it. It does not learn from you and it does not read your health data.
A licensed practitioner is accountable for every clinical decision
This is the rule the whole system is built around. Any output that touches your care — a symptom described, a suggestion, a summary, a piece of patient information — is reviewed and validated by a qualified, licensed healthcare professional before it is acted on or sent to you. That practitioner, and their practice, carry full professional responsibility for it.
AI does not sit between you and your doctor. It sits behind the desk, helping the practice keep up with the messages and the paperwork.
You are told when you are talking to a machine
When you are interacting with our AI assistant rather than a person, we say so at the start of the conversation, in the language you are writing in wherever we can. You should never have to guess.
The same applies to text on this site: where a page has been drafted with AI assistance and concerns your care, a named practitioner has read and signed it before publication.
Emergencies are never left to software
If this is a medical emergency, call 112 immediately, or go to your nearest emergency department. Do not wait for a reply from us on any channel.
Our assistant watches for wording that suggests urgency and hands the conversation to a human as fast as it can, and it will always repeat the emergency number. That is a safety net, not a triage system: it can miss things, and it is never the thing that decides whether your situation is serious.
What happens to what you write
Your messages are processed so that we can answer you and coordinate your care. That is the purpose, and it is the basis on which we hold them, under the GDPR and the Belgian Act on Patient Rights.
Conversations may be used to improve the assistant — but only after they have been de-identified, meaning names, contact details, dates of birth, addresses and other identifying details are stripped out first. We are telling you this because it is true and because the alternative claim, that nothing is ever used, would not be.
What we do not do: we do not sell your data, we do not use it for advertising, and we do not use it to build profiles or scores about you. Your conversations are not shared with your employer, your insurer or anyone else without a lawful basis and your knowledge.
You can ask for a human at any time
Say so — in any language, in any channel, at any point — and a person takes over. You do not need a reason and it will not affect your care. You can also call the practice or ask for a face-to-face appointment instead.
What AI is never allowed to do here
These are not preferences. They are limits written into our Acceptable Use Policy, and they apply to every practitioner and partner on the platform. AI on VeryPatient never:
- Makes a diagnosis, or tells you what is wrong with you
- Chooses, starts, stops or changes a treatment or a prescription
- Decides on its own how urgent your situation is
- Confirms, moves or cancels an appointment without the practice
- Publishes clinical content in a practitioner's name without that practitioner reading and signing it
- Decides anything about your insurance, reimbursement or fitness for work
- Pretends to be a human clinician, or presents its text as written by one
What we deliberately do not claim
We do not claim that your conversations never touch a model, because they do — that is how the assistant answers you. We do not claim they are never used to improve it, because after de-identification they may be. And we do not display an AI safety badge, because no such badge means anything: what matters is which decisions a machine is allowed to make, and on this platform none of them are clinical. If you want to know precisely how a particular answer was produced, ask us and we will tell you.
Something an AI wrote here was wrong. Tell us.
If an answer you received was inaccurate, unfair, confusing or simply felt wrong, we want to see it. A human reads every report, and it is how this page and this system get better. You can also use the "Report issue" button in your patient or practitioner dashboard.
safeguards@verypatient.healthLast updated 2 August 2026