Welcome to the forefront of conversational AI as we explore the fascinating world of AI chatbots in our dedicated blog series. Discover the latest advancements, applications, and strategies that propel the evolution of chatbot technology. From enhancing customer interactions to streamlining business processes, these articles delve into the innovative ways artificial intelligence is shaping the landscape of automated conversational agents. Whether you’re a business owner, developer, or simply intrigued by the future of interactive technology, join us on this journey to unravel the transformative power and endless possibilities of AI chatbots.
If a chatbot misdiagnosed your health problem, it might be your fault. With the cough and cold season coming, a lot of people are likely to fire up their favorite artificial intelligence chatbot to try and diagnose what they’re feeling. Earlier this year, the Kaiser Family Foundation found 1 out of 3 Americans were using AI to help get answers to questions about their health.
The problem is, the answers they get aren’t always going to be accurate. But research from England’s Oxford University suggests the problem isn’t always the reliability of the chatbots, but the person trying to diagnose their own problem.
“When you start using chatbots for medical advice in the real world with people off the street that don’t have sort of medical expertise and background,” said Andrew Bean, a doctoral student at the Oxford Internet Institute. “How well does that work?”
The answer depends on the information those models were given. When doctors wrote out the situation with a patient and fed it into the chatbot, the answers were pretty much on the mark.
“If you gave a model the full context of what’s going on with the patient in a structured written format, has all the details there, most of the time they were actually able to give good advice,” Bean said.
But when a regular person without a medical background was told to provide the information to the chatbots instead, the answers were less reliable.
“What I think it definitely highlights is a gap between how capable the models are and how actually usable they are,” Bean said. “Without medical knowledge, the users just don’t know what the right questions to ask are. They don’t know what the symptoms that they should be describing are, or sort of how the best way to describe them is, and so they need more support from a system they’re interacting with to help them to figure out what those key things are.”
In addition, the users had thought they were getting good answers.
“I think the biggest takeaway would be that you really need to have some sort of trusted way of verifying anything that you get out of a language model (Chatbot),” Bean said. “Users were coming to responses that they actually rated them quite confidently. They felt that they had gotten to a good answer. Yet when we compare to what doctors who had originally written the scenarios said good answers were, they typically weren’t.”
Bean mentioned another study that found most AI chatbots get health questions during hours when doctors’ offices aren’t open, like overnight, which would provide some insight as to why so many people are turning to chatbots for questions about their health. And the technology keeps improving.
“We’re at a stage where maybe a language model is helpful, but you’re going to need to verify it,” he said.
Get breaking news and daily headlines delivered to your email inbox by signing up here.
© 2026 WTOP. All Rights Reserved. This website is not intended for users located within the European Economic Area.
© 2026 WTOP, LLC. All rights reserved. This website is not intended for users located within the European Economic Area.
YouTube Terms of Service | Google Privacy Policy | Google Terms of Service
