Marketers are increasingly seeking assistance from generative AI (GenAI) on tasks that used to be performed by human employees. To benefit from GenAI, marketers need to know how to prompt it. Do marketers adjust their communication style when instructing a GenAI compared to when instructing human assistants? Should they? Five pre-registered studies (plus two supplementary replications) reveal that marketers naturally adapt their communication when instructing GenAI by using fewer filler words and including a higher proportion of task-relevant keywords compared to when instructing human employees. This adaptation is largely beneficial. GenAI produces more persuasive, accurate, and purchase-motivating content when marketers provide concise, keyword-focused instructions rather than more verbose instructions that typically benefit human-to-human communication. Consequently, treating GenAI like a human reduces its ability to generate high-quality marketing content. This research contributes to communication theory by demonstrating that marketers tend to accommodate GenAI using less filler and more information-dense keywords. It contributes to marketing practice by showing that the intuition and advice to speak to GenAI as if it were a human can backfire by leading it to produce less accurate, less persuasive, and lower-quality output.