The Lancet Digital Health’s recent evaluation of LLM susceptibility to medical misinformation is a textbook case of "medical archaeology" that fails to keep pace with the frontier of the field. By the time this study cleared the traditional peer-review cycle, the models it benchmarked—such as Llama 2 and GPT-3.5—were already obsolete, leaving the capabilities of actual clinical frontier models like Gemini 2.0 and GPT-5 entirely unaddressed. We need an expedited framework for medical AI research; otherwise, these journals are merely documenting history rather than providing actionable clinical guidance for the tools currently in use.

thelancet.com/journals/…