This is a four-arm, online randomized controlled trial examining whether conversational interactions with large language models (LLMs) can effectively correct health misinformation beliefs. Participants who endorse the target misinformation belief are randomly assigned to interact with one of three LLMs (ChatGPT, DeepSeek, or Gemini) or to pre-collected snapshot of search engine results. Outcomes include participant-reported belief and satisfaction measures, as well as expert evaluations.