Experts have raised safety concerns for patients seeking medical information over the Internet. This study examined whether a conversational assistant such as Siri, Alexa, or Google Assistant would provide accurate information in response to a medical question. Lay participants queried a randomly assigned conversational assistant about their own health-related question and standardized questions relating to medication use and recognition of symptoms. Conversational assistants were unable to answer the majority of questions. Among the answered questions, a significant proportion of suggested actions (29%) could lead to harm. The authors conclude that conversational assistants are neither safe nor effective in providing actionable medical information.