News and Insights
10 Prompt Engineering Tips for Writers
January 18, 2024
Crafting an effective generative AI prompt takes knowledge and a bit of skill. For those using large language models (LLMs) like ChatGPT, Claude, and Copilot, having a clear sense of the ideal output requires an understanding of the tool, the nuances of its interface, and its limitations. Followed are some tips for getting the most out of your generative AI copywriting experience.
- Have a vision for what you hope to achieve
Successful prompt engineering is largely a matter of knowing what questions to ask and how to ask them. When using an AI tool, define clear goals for the interaction and develop an outline of the anticipated results (i.e., what you want to achieve, what the audience should know, and any associated actions for the system to perform). - Format matters
AI systems are designed for conversational interactions. This means they can work with simple, direct requests using casual, plain-language. In cases where requests are complex, provide detailed, carefully structured prompts that adhere to a form that is consistent with the system’s design. Consider, for example, this prompt format: “Act like” + “write a” + “define an objective” + “define your ideal format”. - Make clear, specific requests
Generative AI tools are neither psychic nor telepathic. They act based on what they can interpret from a given prompt, so form clear, explicit, and actionable requests, understand your desired outcome, and work to describe the task that needs to be performed (or the question that needs to be answered). - Pay attention to prompt length
Most generative AI tools place limits on the number of characters that can be used in prompts. AI tools often struggle to decipher prompts due to the complexity involved in organizing and prioritizing the essential elements of a lengthy request, so break complicated asks into smaller tasks for the best results. - Choose your words with care
AI tools can be precise in their use of commands and language, including not knowing how to respond to unrecognized commands or language. Avoid ambiguity, colorful language, metaphors, and slang – the most effective prompts are those that use clear and direct language. - Ask open-ended questions
Generative AI tools are designed to create – that’s their job! Simple ‘yes’ or ‘no’ queries limit their ability to do so and will likely result in outputs that are brief and boring. Open questions, by comparison, give them room to explore and to deliver responses that are more comprehensive and detailed. - Include context
Generative AI tools can frame their outputs to meet a wide array of writing goals and expectations, from short, generalized summaries to long, detailed articles. Entering a well-crafted prompt that includes context will help your preferred AI system tailor its outputs to your needs and audience (“Explain the basic functions of generative AI for communications professionals”). - Test length goals for outputs
While useful in theory, incorporating guardrails like output length into prompts won’t always yield the intended response. Using keywords like “simplified” and “concise” (versus “lengthy” and “detailed”) may help, but you can expect inconsistent results at best. This is because natural language processing models (NLPs) such as GPT-4o are trained to predict words based on language patterns, not count. Thus, LLMs can usually follow approximate guidance (such as “provide a two- or three-sentence response”), but struggle to precisely quantify characters or words. - Avoid conflicting terms
Prompts including ambiguous or contradictory terms (i.e., “a detailed summary”) send generative AI models conflicting signals about the expected depth and length of an output. The most effective prompts use positive and not negative language. Keep in mind that AI models are trained to perform specific tasks, so asking one not to do something is meaningless, unless there’s a compelling reason to include that exception in a parameter. - Use punctuation
Just as humans rely on punctuation to help parse text, AI prompts can also benefit from the use of commas, quotation marks, and line breaks. These help the systems understand and respond to prompts appropriately.