The answers provided are not always correct or up to date. And be careful about privacy
THE’artificial intelligence (AI) is one of the fastest-growing technologies in human history, in terms of investment, adoption and technical capacity. And more and more often, people are turning to it to search medical information or receive an initial orientation on symptoms, diagnosis And treatments. Unlike an old-fashioned query on a search engine, which returns generic results or those referring to other people’s cases, a chatbot is able to provide highly personalized responses. And not only because they are built on the individual case described by the user, but also because of the formulation, calibrated to the preparation of the reader. A targeted approach which however raises major doubts: how much we can really trust AI responses when we talk about health?
AI and medical diagnoses, the advantages
—
One of the main strengths of medical chatbots is theaccessibility. Anyone with an internet connection can get basic information about symptoms, therapies or prevention strategiespractically at no cost, from the comfort of your own home and without appointments. A convenience that is also useful for investigating an already diagnosed problem or for better understanding the mechanisms of a therapy. People also perceive a greater sense of privacy and they feel freer to talk about delicate ailments or symptoms compared to a face-to-face meeting. Added to this is the innate ability of the language models of translate complex medical concepts in a language accessible to all; a real godsend, for the understanding diagnoses or reports. Some recent research also points to a possible role of support between one visit and anotherfor example to remember appointments or to monitor the progress of a therapy. In some cases, AI can even offer a rudimentary psychological supportsuggest relaxation techniques or strategies for dealing with anxiety and stress. But each of the pros comes with a series of cons.
The risks of AI for medical consultations
—
The main disadvantage of AI in the medical field is the lack of guarantees onaccuracy of the results. The quality of the information provided by a chatbot depends greatly on the quality of the data with which it was trained; data that does not always come from verified or updated scientific sources. In “generalist” AI there is not always a review on this aspect. This can lead to the formulation of inaccurate or completely incorrect answers. Added to this is the great flaw of LLMs, namely the hallucinations. The answers, often formulated in a complex and authoritative way, are made to seem credible, to emulate an authoritative response, which can generate a false perception of security. Unlike a doctor, AI cannot interpret personal clinical data, cannot analyze images or diagnostic tests and does not possess empathic or contextual skills. Furthermore, it can perpetuate bias and distortions deriving from the sources on which it was trained. So how do you regulate it?
AI in medicine, to trust or not?
—
Establishing the degree of reliability of a medical chatbot is not a simple matter. Trust in the healthcare sector is normally built in direct relationships with doctors and professionals, where listening, empathy and competence count. In the case of artificial intelligence, the parameters of judgment are more nuanced and broader. According to 2025 research, the accuracy of ChatGPT’s medical responses varies between 20% and 95% depending on the type of questionwith more reliable results in general situations and less precise in specific cases. The authors of the study concluded that you should not use ChatGPT alone to make a medical diagnosis. The most reliable answers are obtained by comparing the information provided by the AI with that coming from accredited sources, such as health institutions, medical associations or scientific journals. In any case, to correctly evaluate symptoms, medical history and therapies, the a doctor’s opinion remains essential. And for reasons of privacyit is also important to avoid feeding entire medical records to a chatbot. Better to limit the data to what is strictly necessary.
Practical advice for using AI
—
If you decide to use artificial intelligence as a support in healthcare, it is necessary be specific and formulate clear and detailed questions, to limit the possibility of error. It is also helpful to always provide context, describing symptoms, habits or medications taken. It’s not a bad idea to ask the chatbot to check the references to the studies or publications and then check their validity on official databases and respectable sites, such as PubMed. Finally, the most important rule of all: AI does not replace a doctor.
© ALL RIGHTS RESERVED