Symptom Search Shifting from Google to ChatGPT? The Assurance and Risks of AI Health Consultation

Symptom Search Shifting from Google to ChatGPT? The Assurance and Risks of AI Health Consultation

One in Four Turn to AI Before Doctors: The New Norm in Health Consultation and Its Risky Boundaries

Feeling unwell? A sore throat, a fluttering chest, insomnia, or unfamiliar terms in a medication leaflet? Traditionally, many would first turn to a search engine. However, this initial step is quietly shifting. Instead of typing keywords into a search bar, more people are now engaging with chat interfaces, asking, "What could this symptom mean?" "Should I see a doctor?" "Can you explain the side effects of this medication in simple terms?"

A survey by EY, highlighted in German media, reveals this shift in numbers. In Germany, 54% of respondents have used AI applications in recent months, with 23% using AI to quickly obtain health information without visiting a doctor. Additionally, 20% reported analyzing their symptoms using chatbots like ChatGPT. Some 14% used it for mental health discussions, while 10% used it to understand medication leaflets.

The key point here is that AI is no longer just a "convenient search tool." Users are turning to AI not only for physical symptoms but also for concerns about anxiety, loneliness, medication doubts, and whether to seek medical attention. Instead of comparing search results themselves, they ask AI, "What should I consider in my case?" This is more akin to consultation than mere information gathering.

Underlying this trend are issues with medical access: difficulty in securing hospital appointments, limited consultation hours, concerns about costs and waiting times, and uncertainty about the severity of symptoms. In this context, AI, which provides responses 24/7, appears attractive, especially in situations where people want to quickly alleviate anxiety, feel their issue might not warrant a doctor's visit, or wish to organize their thoughts before a consultation.

Reactions on social media reflect this convenience. Public posts and forums feature positive comments like, "I can organize my questions before visiting the hospital," "It helps me create notes to explain to the doctor," and "Entering my symptoms led to a recommendation to seek medical attention, ultimately prompting a hospital visit." On Reddit, experiences where AI helped organize differential diagnoses or assisted in deciding whether to seek medical attention frequently become topics of discussion. Some posts even mention that AI responses prompted consultations with doctors, revealing overlooked possibilities.

Conversely, there is also strong caution on the same social media platforms. Concerns include "AI is not a doctor," "It’s dangerous because it gives plausible answers," "Anxious people might trust AI's assertive tone," and "More people might end up not going to the hospital." These concerns are not merely emotional. Health-related decisions involve many factors, such as age, medical history, medication, test values, living environment, symptom changes, and the individual's condition. The few lines of information entered into a chat cannot match the quality and quantity of information a doctor gathers in a consultation room.

AI excels at organizing general information, listing possibilities, and simplifying complex terms. For example, explaining the meaning of test items, rephrasing medication leaflets in simple terms, listing questions to ask during a consultation, or summarizing symptom progression chronologically. When used this way, AI can aid patient understanding and improve communication with doctors.

However, AI struggles to determine if "this person is truly in danger right now." Symptoms requiring urgent attention, such as chest pain, shortness of breath, severe headaches, changes in consciousness, sudden paralysis, risk of self-harm, or severe allergic reactions, cannot be safely assessed through text alone. If AI provides a response that could be interpreted as "it's okay to wait," it might lead to delays in seeking medical attention. Conversely, overemphasizing danger could amplify anxiety.

In the mental health domain, even greater caution is needed. AI does not reject, responds immediately, and listens for extended periods. For those feeling lonely or anxious, this responsiveness can feel like a lifeline. Social media also reflects sentiments like "I can tell AI things I can't tell anyone else" and "It helps when I feel anxious late at night." However, effective mental health support requires the ability to detect crisis signs, maintain continuous relationships, connect with local support systems, and involve professional judgment. The ability of AI to respond empathetically is not equivalent to providing responsible support.

An intriguing aspect of the survey is that Germany's usage rate is lower than the global average. Globally, 62% of people use AI for health-related purposes, with India at 90% and China at 89%. Meanwhile, Japan's rate is 42%, one of the lowest among surveyed countries. Germany's rate is also below the global average at 54%. This is influenced not only by the degree of technological adoption but also by trust in the healthcare system, perceptions of personal data, cultural distance from AI, and expectations of regulation.

In Germany, despite the widespread use of AI, only 28% of people said they "understand AI well and can use it meaningfully." This indicates that while many use it, they may not fully understand its mechanisms or limitations. This gap is the greatest risk in health consultations. AI responses are natural, calm, and sometimes appear expert-like. However, the smoothness of the text does not guarantee accuracy.

 

The repeated observation on social media that "AI's answers are too confident" highlights this issue. In situations where a human doctor might say, "We need tests to know," "It's possible but not certain," or "Better to see a doctor just in case," AI presents possibilities in organized text. This clarity provides reassurance but can also lead to false confidence. If users decide, "AI said it's okay," the convenience turns into a risk.

So, should AI health consultations be avoided? The answer is not simple. In reality, many people are already using them. Prohibition or denial alone cannot keep up with the actual usage. What is needed is to position AI not as a replacement for doctors but as a "supplementary line before and after reaching the doctor."

For example, organizing the progression of symptoms before a consultation. Making bullet points of what to convey during a consultation. Understanding the general meaning of technical terms in medication leaflets. Considering questions to confirm with the doctor about test results. Learning about general options for lifestyle improvements. Used this way, AI can become a tool that enhances patient autonomy.

Conversely, there are uses to avoid. Adjusting medication based solely on AI responses. Canceling medical appointments. Presuming a diagnosis. Judging emergency symptoms through chat alone. Relying solely on AI for severe mental health crises. Entering identifiable medical information carelessly. These actions carry more risk than convenience.

The medical side also needs to change. Assuming patients will research with AI, doctors and medical institutions should not dismiss AI use but explain "how to use it safely." When patients bring AI responses, merely denying them will not build trust. Instead, it's important to collaboratively verify which parts are valid and which are risky.

Companies and AI developers also bear responsibility. In health consultations, it's essential not only to make answers easy to understand but also to clarify limitations, promote seeking medical attention in emergencies, protect personal information, involve expert supervision, and suppress dangerous advice. Especially in areas related to mental health, medication, diagnosis, and treatment plans, AI should be designed to prevent users from mistaking it for a "safe expert."

This survey does not suggest that AI will suddenly replace healthcare. Rather, it highlights how much anxiety and uncertainty people experience before accessing healthcare. AI is chosen not because it is perfect but because it provides quick responses, is easy to ask questions, allows for embarrassing inquiries, and lowers the psychological barrier before visiting a medical institution.

However, the true goal of health consultations is not for AI to provide answers. It is for those in need to connect with appropriate medical care and support at the right time. AI might widen the entrance, but if the entrance becomes the exit, it becomes dangerous.

The behavior of "asking AI first" will likely become more common. Therefore, the question is not whether to use AI but where to draw the line between relying on AI and connecting with human experts. In the era of AI health consultations, what is needed is neither overconfidence in nor rejection of AI. It is the calmness to use AI as a convenient consultation partner while not separating the final judgment from human medical care.


List of Source URLs

Related Articles