Diabetes AI Wins Expert Ratings Over General Chatbots

Managing diabetes is a complex job. It is not just about one type of medicine. You have to mix Western treatments, older healing styles, and daily habits. Computers are getting better at writing health advice. But big chatbots often guess. They do not always check their facts. This makes doctors nervous about trusting them blindly.

A new system called CDTD-Agent was built to fix this gap. It does not just guess. It looks inside a special library first. That library holds twelve books, eight rule guides, and 183 expert-checked patient stories. Experts checked those stories before they were added. The system pulls the right info, thinks about it, and then gives an answer. This helps it understand different medical styles at once.

To see if it works, thirty senior doctors tried it out. They did not know which advice came from the machine. They looked at eight different things, like safety and whether the tests made sense. The computer advice was compared against famous chatbots and two real doctors. The experts gave scores for every single option. But remember, this was just one patient case.

The numbers showed a clear winner. The new system scored much higher than the general chatbots. It won in five areas, including safety and treatment choices. The biggest gap was in Western medicine plans. The machine got a 4.07 score while the others struggled between 1.37 and 1.60. Even when the team tested the basic model without the smart search, the full system still did better.