Study by CSIRO and The University of Queensland highlights how the phrasing of prompts influences the accuracy of health information provided by ChatGPT, showing a significant variance in effectiveness based on prompt design.