This study examines the frames in natural disaster-related news in Madagascar and Malawi. Previous studies have shown that the media would mainly address conflict in their coverage. However, practices such as solutions journalism have recently emerged and have proven to be efficient in terms of engagement. There could therefore be a shift in news frames as solution-based stories are also increasing, and this study aims to observe whether or not there is a change in the framing of news. Using ChatGPT to do the frame analysis, this study focuses on the tropical cyclone Freddy that hit Madagascar and Malawi in early 2023. A total of 329 articles from the mainstream print media of both countries were retrieved and analyzed in this study. The findings show that coverage of natural disasters still tends to focus on the response and the extent of damages. On the other hand, ChatGPT was proven to be 75.69% accurate in terms of frame analysis but could however not clearly distinguish between the response and solution frames. This study suggests, however, that proper prompting on ChatGPT could lead to better differentiation. The implication of the results for computational content analysis was also discussed.