Just as attackers have used SEO techniques to poison search engine results, they could rinse and repeat with artificial intelligence and the responses LLMs generate from user prompts.