The internet is quite hypochondriac and chatbots like ChatGPT are too, GP Tobias Bonten regularly tells his patients in the consultation room. They often bring a point-by-point, “precisely formulated” list of questions that “has been made very clear by ChatGPT,” says Bonten – in addition to being a general practitioner in Katwijk, he is also an associate professor at the Leiden University Medical Center (LUMC) and a researcher in digital care. Then he explains to his patients that the internet contains more stories about very serious illnesses than stories of people who recovered from a condition on their own.
Doctors know that many patients consult artificial intelligence in the form of an AI bot like ChatGPT. Self-management is important for many people and reliable information is essential, but not always available. Commercial chatbots are known for providing incorrect medical information and storing a lot of (sensitive) data about their users.
Yet general practitioners also see advantages from the rapid advance of AI in practice. They also like to make use of technological developments. Especially with simple tasks: administration, looking up medical information, requesting complaints digitally and when contacting patients. Whether this saves doctors time has not yet been proven. The developments certainly make the work more fun, says Bonten: “I can pay more attention to the conversation with the patient. I have to click and type less.”
Dutch AI alternative
GPs are not necessarily against patients looking up information on the internet. But then that information must be reliable. To generate answers, commercial AI bots scan almost the entire internet and with all that information the computer model always generates an answer – even if the information does not correspond to Dutch guidelines for diseases. And if the answer is simply not known, the bot ‘hallucinates’.
The popularity of AI bots in the consulting room reminds Bonten of the rise of Google. This also allowed patients to find a lot of information about their health. But chatbots make it even easier for patients to search for health information. Bonten: “An important difference between chatbots and Google is that the chatbots pleasers are. They always want an answer, and sometimes come up with a source. If you have a stomach ache or are tired, that does not immediately mean that you have cancer.”
The Dutch Thuisarts hopes to offer an alternative to commercial AI chatbots. Last Thursday, the platform therefore launched its own AI assistant on the existing website. The chatbot is only trained with information from its own website and refers users to information that matches the complaints described. The assistant is not allowed to give medical advice and, according to Thuisarts, does not collect sensitive personal data. The goal is that people can “make well-informed choices themselves” and collect information in an “interactive way,” says Thuisarts director Swanet Woldhuis.
Thuisarts is used by many general practitioners. It helps doctors convey information to patients. The website has reduced the number of GP consultations over the past ten years, GP Bonten knows from previous research he has conducted. Thuisarts has approximately 72 million visitors per year and is an independent, non-profit platform with reliable health information. The information on Thuisarts is created by doctors and tailored to Dutch guidelines.
The home doctor had to test the AI assistant against many rules. On Tuesday it emerged that the website Moetiknaardedokter.nl (MINDD) does not comply with strict European regulations for online healthcare applications. The NOS wrote this on Tuesday based on its own data research. It is not clear what the tool bases its advice on, according to NOS, and that is a problem. The supervisory authority, the Healthcare and Youth Inspectorate (IGJ), also confirms this.
Also read
A Dutch doctor prefers not to let the AI ethics of big tech impose on him

Digital portal for questions and recipes
The government wants to help doctors and patients in the coming years with ‘digitalization of healthcare’. She wants to invest 400 million euros in both 2027 and 2028 to further implement AI applications in healthcare organizations as quickly as possible. Various apps, websites and programs are being developed that will allow patients to do more from home and save doctors time. Patients often have an online patient environment with their GP, where they can request (repeat) prescriptions or ask questions – without a call or a visit to the practice. Sometimes patients can also upload photos of a condition into the system so that a GP can assess it from a distance.
Mascha Bevers’ general practice in Amsterdam is a textbook example of ‘digital care’: Bevers has integrated AI into its computer system. “When I call someone, my information system immediately sees which patient I am on the line. I can open their file, record the conversation and have it summarized.” Bevers then manually checks whether the summary contains the essence of the conversation with the patient and can supplement the report with things that she considers important. For example, emotions in a patient’s voice, which AI cannot pick up.

Mascha Bevers from Huisartsen Westerpark.
Photo Saskia van den Boom
Bevers is a ‘flexible doctor’: “What the assistant normally does on the telephone, our patients can do all themselves in the digital system.” The practice’s 3,200 patients have a personal digital ‘portal’ where they can request repeat prescriptions, ask questions, view test results and make a telephone appointment. Bevers: “We can immediately solve 70 percent of the questions digitally. What remains we see in practice, just like emergencies.”
It is not known exactly how many general practitioners use AI. What is known is that 39 percent of general practitioners indicated in 2025 that they would not use AI applications. This is evident from the Monitor Digital Care from 2025, carried out by the RIVM, Nivel research institute, and the National eHealth Living Lab.
So far, digitalization has not directly led to a lower workload for general practitioners. Jelle Keuper, researcher in digital care in general practices at Nivel, knows this. “We do not yet know whether the number of consultations or telephone calls actually increases or decreases due to AI.” Keuper does know that “about half of general practitioners indicate that they are (very) positive about the influence of AI on workload.” “The same applies to job satisfaction and the quality of care.”
At the same time, about a quarter of general practitioners are (very) negative about the influence of the use of AI on workload, the researcher says. Critics of AI use in healthcare warn, among other things, about privacy risks if patient data is collected and stored for digital healthcare applications. Data leaks at healthcare organizations are not unthinkable, as recently became apparent at healthcare software company Chipsoft. Several GP practices were also hacked.
Using digital applications is also not easy for all GPs, says GP and researcher Tobias Bonten. “The biggest stumbling block is the connection of new applications with existing general practice systems.” Consider different login details for different systems, and copying information from one program to another. The popular transcription program Juvoly summarizes the conversation between patient and GP, but Bonten must always check the result carefully, he says: “Sometimes the AI bot writes things that are not correct.”
Do not store personal information
Both Bonten and Bevers are positive about the idea of a chatbot from Thuisarts. Bevers: “GPs always advise Thuisarts in conversations, so that patients can fall back on reliable information. If an AI bot can be closely monitored for hallucinations and can help patients find good information more easily, that is fantastic.”
“We do everything we can to provide reliable information to our users,” says Thuisarts director Woldhuis. The new chatbot is programmed not to store any personal information, she says. If someone does share their name, address or lifestyle with the chat, Thuisarts “removes that personal information.” Woldhuis: “We collect data about questions and answers to train the bot to provide correct information, but no personal data, including IP addresses.” The data is stored on a rented, secure server within Europe, according to Woldhuis.

The AI assistant is well secured, says Woldhuis: “We have had our tool independently tested against the EU AI Act and the legislation for medical devices.” The Home Doctor Assistant uses existing technology from external providers, including Google Gemini. Thuisarts is therefore unable to provide complete transparency about the software.
For the time being, the AI assistant remains available on the website for a relatively small group of Thuisarts visitors. The assistant is still in a testing phase, only 1 to 2 percent of website visitors can now use the bot. To make the assistant available to all visitors, Thuisarts says it needs more money. Thuisarts also wants to integrate the assistant into the Thuisarts app for mobile phones.
The Thuisarts AI assistant is not allowed to trick users, and is so strictly programmed that in the initial phase it could not even respond to a simple ‘Good morning’, says Woldhuis. “That is not on Thuisarts.nl,” was the answer from the chatbot at the time. That has now been remedied. The answer is now: “Hello! I will help you find information on Thuisarts. What subject do you want to know about?”
Also read
AI helps the GP, but time savings? There isn’t, is the honest answer


