A small digital marketing agency uses simple prompts with a publicly available AI chatbot with web search, such as Copilot or Bard, to generate initial drafts for short blog posts. However, when the agency prompted the chatbot to “write a blog post on balanced diets,”  it responded surprisingly quickly with a draft that contained made up, or "hallucinated," information. For instance, the AI wrote that "A balanced diet can make you invisible to mosquitos."

How can the prompt be improved to reduce the chance of hallucinations?