Artificial intelligence is increasingly becoming the first place people turn when they notice an unusual symptom, receive test results, or simply try to understand what is happening in their body. ChatGPT and similar services answer quickly, in a structured way and in plain language, creating the illusion of a full medical consultation. However, as family doctor Lilia Sarakhman emphasizes in an exclusive comment for Wave RBC.UA, using a neural network as a substitute for a doctor is dangerous: even a highly convincing algorithmic answer cannot replace a proper medical examination.
Why AI 'Doesn't See' the Patient
The key problem, according to Lilia Sarakhman, is that artificial intelligence does not see the patient. During a personal examination, a doctor assesses far more than a dry description of symptoms: they pay attention to skin color, the nature of breathing, the presence of swelling, the characteristics of pain, the body's reaction during the exam, and many other details. 'Sometimes a single detail observed during the exam completely changes the diagnostic and treatment approach,' the doctor notes. That is why even a very detailed description of your condition in a chat does not always allow for a complete clinical picture, on which sound treatment is based.
The Confident Tone as a Trap for the Patient
The second, no less insidious problem is that an artificial intelligence error can look like a perfectly professional answer. ChatGPT is able to formulate information confidently, structurally and clearly, so for a person without medical training it can be extremely difficult to determine where reliable information ends and an error begins. As a result, a patient may believe a misinterpretation of symptoms, feel reassured, and postpone seeing a doctor. This is especially dangerous when, behind what seem to be minor symptoms, there is actually a condition requiring emergency care.
When a Symptom Requires Urgent Assessment
The doctor reminds us that symptoms such as chest pain, shortness of breath, dizziness, weakness, or a headache can have dozens of different causes. Some of them are indeed relatively harmless, but some require urgent medical assessment. It is in such situations that 'home' diagnosis with a chatbot turns from a convenient tool into a risk factor: a person spends time discussing things with an algorithm instead of seeking live help.
How to Use AI on Health Issues Correctly
At the same time, artificial intelligence itself, according to Lilia Sarakhman, is not the enemy. It can be a useful tool if used as intended: with AI you can break down an unfamiliar medical term in plain words, prepare a list of questions for your doctor, structure information about your own symptoms, or better understand the recommendations you have already received. 'A person's health is certainly not a case where you should test how well an artificial intelligence algorithm works,' the doctor concludes. In other words, the neural network works as an assistant and a 'translator,' but not as a diagnostician or treating specialist.
Legal Precedents: When AI Advice Turns Into a Lawsuit
The danger of replacing a doctor with an algorithm has already gone beyond theoretical warnings. As reported by Russian media, in July 2026 a former pastor filed a lawsuit against OpenAI, claiming that ChatGPT advised him not to see a doctor, which, he says, harmed his health. Such precedents clearly demonstrate that a 'confident' neural network answer can not only contain an error but also directly contradict the basic principle of the need for in-person medical care. Against this backdrop, statements about the rapid growth in the accuracy of diagnostic neural networks (including reports that certain AI models outperform the diagnostic decisions of major corporations) sound like an argument 'for' developing the technology, but not an argument 'against' a live doctor.
Contradictory Data
In the public discourse around medical AI, two polar positions coexist. On the one hand, developers and researchers emphasize that diagnostic neural networks are achieving high accuracy, and some models, according to media reports, outperform major technology companies on certain metrics; AI is positioned as a powerful tool capable of easing the burden on the healthcare system. On the other hand, practicing doctors such as Lilia Sarakhman insist that no model sees the patient, takes the context of the examination into account, or may present an error in a professionally sounding form, and that lawsuits (such as the case of the former pastor and OpenAI) show real harm from blindly trusting a chatbot's advice. Both viewpoints reflect different aspects of the same reality: the technology is indeed developing and is useful as an auxiliary tool, but its current limitations make it impossible to use AI as a replacement for an in-person medical consultation. An honest reflection of this contradiction is important so that the user does not perceive the neural network either as a panacea or as a useless toy, but as a limited assistant.
The bottom line for the reader is simple: ChatGPT and similar services can help you make sense of terms, formulate questions, and structure symptoms, but the decision on diagnosis and treatment remains the prerogative of a doctor who sees the patient as a whole. In matters of health, as Lilia Sarakhman warns, you should not be testing how well an algorithm works — here a person's health and life are at stake.
