Artificial Intelligence (AI) chatbots have become an increasingly popular tool for answering health-related questions and providing information. However, concerns about biased or inaccurate information persist, prompting developers to continuously update and improve chatbots‘ ability to handle health misinformation. Over the past year, advancements in AI technology have enabled chatbots to better identify and correct misinformation, cross-referencing information from multiple reliable sources to verify claims and detect inconsistencies.
Recent updates from major players in the AI chatbot space, such as Google and Microsoft, reflect this trend towards enhanced accuracy and reliability. Google’s Bard has been rebranded as Gemini, and Microsoft’s Bing Chat is now known as CoPilot. OpenAI has also upgraded ChatGPT, introducing new features like real-time voice interactions. These improvements aim to provide users with more trustworthy information and a better user experience while interacting with AI chatbots.
To assess how well-known AI chatbots have evolved in handling health-related queries, KFF conducted a series of tests with ChatGPT, Google Gemini, and Microsoft CoPilot. The tests involved asking the chatbots about false claims related to health, monitoring their responses over time to evaluate their accuracy and reliability. While each chatbot generally identified false claims, they varied in their directness and approach to addressing them.
For instance, ChatGPT initially took a more cautious approach, highlighting the complexity of certain issues and suggesting the need for further research. Over time, it became more assertive in labeling claims as false. Google Gemini and Microsoft CoPilot, on the other hand, tended to directly refute false claims. Interestingly, the chatbots also differed in how they cited sources to support their responses, with some providing specific studies as evidence while others offered more general references.
In terms of referencing public health institutions, the chatbots showed evolution over time. ChatGPT initially cited specific agencies like the CDC or FDA for COVID or vaccine-related questions, while Gemini and CoPilot referenced trusted sources consistently. By 2024, ChatGPT began referencing a broader range of health topics and specific institutions, while Gemini shifted to providing general resource links and CoPilot maintained consistency in citing public health organizations.
In conclusion, while AI chatbots can be a convenient tool for accessing health information quickly, they are not without their limitations. It is essential for users to be aware of potential inaccuracies, misrepresentations, or gaps in information provided by chatbots. Double-checking answers from multiple sources and staying informed about updates to AI chatbot systems are recommended practices to ensure reliability and accuracy when using these tools. By understanding the advancements and challenges in how AI chatbots handle health misinformation, users can make more informed decisions about the information they receive from these platforms.
