LETTERS TO THE EDITOR IN MEDICINE SHOULD NOT BE GENERATED BY AI

Many academics might be enthusiastic about the current wave of generative artificial intelligence (GAI), using large language models like ChatGPT or DeepSeek. We show that the reputation of letters to the editor (LTEs), which play important communicative functions in biomedicine recognized by several authors as well as the ICMJE recommendations, is being damaged by the abuse of GAI. There has been a recent surge in LTEs generated by the undeclared use of AI and passed off as human-written. We consider that LTEs should be a forum for human engagement that retain at least two different functions in the age of GAI: to question published conclusions and to generate new ideas. Both of these are essential skills for researchers and clinicians, particularly in their early career. We encourage academics to protect the writing and reading of human LTEs as a forum for human engagement with published research, the empirical facts and the values of medicine. We consider that GAI should only be used for secondary purposes such as editing and translation rather than text generation.