A new study has found that using AI for medical advice does not help people make better health decisions. The research was published in the journal Nature Medicine.
The authors said more people now rely on AI and chatbots to check symptoms. However, there is no strong evidence that this method is safer than traditional sources.
Researchers from the University of Oxford worked with doctors to create ten medical scenarios. These ranged from minor illnesses to serious conditions like brain bleeding.
The team first tested three major AI systems without human users. These included OpenAI’s GPT-4o, Meta’s Llama 3, and Cohere’s Command R+.
In these tests, the AI systems correctly identified medical conditions in nearly 95 percent of cases. They also suggested the right action, such as calling an ambulance, in about 56 percent of cases.
The next phase involved 1,298 participants in Britain. Some used AI tools. Others used search engines, the NHS website, or personal experience.
When real users interacted with AI, results were much weaker. Correct conditions were identified in less than 35 percent of cases. The right action was chosen in less than 45 percent.
These results were no better than people using standard internet searches. This showed that AI did not offer a clear advantage.
Adam Mahdi, a co-author from Oxford, said there is a large gap between potential and real use. He explained that people often provide incomplete information.
The study also found that AI sometimes gave misleading advice. In one example, two people described similar serious symptoms. One was told to go to the hospital. The other was advised to rest.
Researchers said both human input and AI responses can lead to errors. This makes AI unreliable for health decisions without professional guidance.
The team plans further studies in other countries and languages. They want to see if results change over time.
The research was supported by Prolific, Dieter Schwarz Stiftung, and both UK and US governments.
In other news read more about Google Plans Space-Based Data Centers for Artificial Intelligence
Experts advise that AI should not replace doctors. It may assist with information, but it cannot ensure safe medical decisions.
