Abstract Text‐based weather forecasts play a critical role in translating complex meteorological information into actionable guidance. Recent advances in Large Language Models (LLMs) raise the question of whether parts of this communication task can be supported by automation. We examine how LLMs generate forecast text, where they perform well and what challenges remain. Proof‐of‐concept studies demonstrate that LLMs can produce coherent forecast narratives from meteorological inputs but may also generate fluent text containing misleading details. Near term use is most likely in supervised settings, with broader adoption contingent on demonstrated reliability and robust accountability.
Gordon et al. (Mon,) studied this question.