When to call your GP? Don’t let ChatGPT decide

It’s evening. You type a complaint into ChatGPT or an AI-powered search engine, and your screen advises you to call the out-of-hours GP line—just to be safe. Dutch urgent care lines say patients are calling after seeing advice like this. At the same time, they’re seeing more care requests and longer waits. But here’s the catch: no one knows how many of those extra calls are actually triggered by AI. What do you do when the chatbot and your gut don’t agree?

In a report by AD and De Stentor, staff at GP out-of-hours services say callers sometimes explicitly mention a chatbot. The paper also ran its own test showing an AI in a search engine quickly pointing people to urgent care. A spokesperson for Neo Huisartsenzorg, active around Nijmegen and Boxmeer, notes that waits can hit twenty to thirty minutes at peak times.

There are solid national figures on usage. According to the notes to industry group InEen’s 2025 GP Urgent Care Benchmark, recorded patient contacts rose 2 percent last year. Average wait time for non-urgent calls rose 14 percent to 5 minutes 30 seconds. The underlying benchmark shows that 74 percent of triage-handled cases were low urgency.

That 74 percent doesn’t mean three-quarters called “for nothing.” If you’re worried, you can’t always judge severity yourself. That’s why triage exists: a trained professional asks follow-ups and decides what’s needed. Also, InEen doesn’t log causes like “advised by ChatGPT.” So you can’t pin that 2 percent rise on AI. The organizations cited in the report also say multiple factors are at play.

A general-purpose chatbot doesn’t see you, can’t examine you, and may not know what you left out. So its advice can skew cautious: consult a doctor to rule out the worst-case scenario. Sometimes that’s right. Sometimes it nudges non-urgent problems toward urgent care. A confident answer is not the same as a triage decision.

AI Wereld previously covered the limits of the claim that ChatGPT Health can reason better than a doctor. Writing strong answers to text-based medical questions is not the same as diagnosing a person or judging urgency. So the real question for a patient isn’t just whether the answer sounds plausible, but which parts of your situation the software could actually assess.

Even the Thuisarts AI assistant, which helps find and explain information from the official GP site, explicitly says it can’t decide if you need urgent help. Thuisarts calls it a tool for understanding general information, not a healthcare replacement. That distinction matters even more when a general chatbot—or an auto-generated answer at the top of a search page without a reliable source—seems to draw a personal conclusion.

The Dutch Thuisarts guide to urgent care gives clear routes by situation. For life-threatening emergencies, call 112. If you think you need a doctor soon and your practice is open, call your GP. In the evening, at night, or on weekends, contact the out-of-hours GP line for issues that can’t wait. Not sure? Look up your specific complaint on Thuisarts—each one explains when to call.

Many out-of-hours services also offer a digital questionnaire. InEen’s benchmark reports that 98 percent of GP cooperatives provide some form of digital self-triage. The Moet ik naar de dokter? questionnaire is a well-known example. That’s different from chatting with a general AI: the questions are designed to assess urgency. Even so, a questionnaire result isn’t a diagnosis—and shouldn’t stop you from calling if you’re worried or getting worse.

You also don’t need to “negotiate” with a chatbot until it says “no urgency.” If you’re seriously worried or your symptoms change, a professional can ask questions your initial search didn’t include. If you do call the out-of-hours line, you’ll typically speak to a triage nurse first. According to Thuisarts, that can lead to self-care advice, a call with a doctor, an appointment, or other help. The phone call is an assessment—not an automatic appointment.

Typing a health question into an AI tool may share sensitive data. A general query rarely needs your name, date of birth, address, or full medical history. Also check what data the service stores and why. We previously explained how private your health data is with ChatGPT. A tool’s privacy policy is separate from whether its medical explanation is accurate—both matter.

What out-of-hours services are seeing mainly shows that AI answers are now part of the care journey. The next step for an honest debate is to measure how many people actually call because of an AI prompt, with which complaints, and what follow-up they receive. Until we have that data, “AI is swamping GP urgent care” is a stretch. For individuals, the line is simpler: use online info to refine your question—and leave urgency decisions to the healthcare channels built for them.