And even if it is true, an AI agent cannot complete the information it provides with the fact that the doctors acquired by experience, explains the doctor of fertility Jaime Knopman. When patients in his Midtown Manhattan clinic provide information from AI chatbots, this is not necessary, but what the LLM suggests may not be the best approach for a patient’s specific houses.
For example, when you consider IVF, couples will receive grades for the forability of their embryos. But ask Chatgpt to try recommendations on the next steps according to these scores alone do not take in other important factors, says Knopman. “This is not only a question of the note: there are other things that enter it” – as when the embryo has been biopsied, the state of the patient’s uterine lining, and where they had later in the past of the AP with fertility. In addition to her years of training and medical education, Knopman says that she “took care of thousands and thousands of women”. This, she says, gives him real information on the next steps to continue that the LLM is missing.
Other patients will be in some way in which they want to transfer embryos, based on an answer they received from AI, says Knopman. However, although the method they have been suggested can be common, another court of action may be more suitable for the specific situation of the patient, she said. “There is science, which we are studying, and we learn to do, but in the art of the reason why a treatment or a treatment protocol is a patient than another,” she said.
Some of the companies behind these AI chatbots have built tools to combat concentrations on the medical information written. Openi, the parent company of Chatgpt, Announcement On May 12, he launched Healthbench, a system designed to measure AI’s ability to answer health questions. Openi says the program was built using more than 260 doctors in 60 countries and includes 5,000 simulated health conversations, Bethaeen users and AI models, with a rating guide designed by doctors to assess the responses. The company says that it found that with the previous versions of its AI models, the course of doctors improves the responses generated by the chatbot, but claims that the last models, available in April 2025, such as GPT-4.1, were as good as the human doctor.
“Our results show that large languages have improved considerably over time and Almedy experts surpass answers to the examples tested in our reference,” said Open IA on his website. “However, even the most advanced systems still have a room for substantial improvement, in particular in the search for a harmful context for sub-special requests and the worst riviability.”
Other companies build specific health tools specially designed for health professionals. Microsoft says that it has created a new AI system – called the May diagnostic orchestrator (May -Dxo) – which by testing the patient diagnosed four times as precisely as human doctors. The system operates by questioning several models of high -speaking large language – the GPT of inletai, Gemini de Google, Claude d’Anthropic, Meta’s Llama and Xai’s Grok – in a loose way that imitates human experts who work together.
New doctors will have to learn to use these AI tools as well as patients who use them, explains Bernard S. Chang, dean of medical education at the Harvard Medical School. This is why his university was one of the first to offer to the student classes On how to use technology in their practices. “This is one of the exciting things of Mons that is happening right now in medical education,” said Chang.
The situation reminds Chang when people started to turn to Internet for medical information 20 years August. The patients would do it like him and said, “I hope you are not one of those doctors who use Google.” But as the search engine has become omnipresent, he wanted to respond to these patients: “You should go to Doctor Who did not do it.” He sees the same thing that happens now with AI. “What type of doctor practicing at the forefront of medicine and does not use this powerful tool?”
