Artificial intelligence (AI) and its related tools, like generative AI, have become integral to performing everyday work for many professionals. As generative AI advances, its ability to produce high-quality and coherent content continues to grow. However, even the most advanced generative AI models out there will perform poorly if they’re not provided with high quality written prompts. To help counter this, here are some tips for effective prompt writing for generative AI: Effective prompt writing for Generative AI.

Use clear language

The most effective prompts are not complex and are written to control an AI model’s understanding about what you want produced.

• Use short sentences
• Use familiar language and avoid complex technical jargon, vague terms, and slang
• Stick to descriptive writing and don’t assume the model’s familiarity with cultural differences

Be specific, but not too narrow

A vague prompt will leave any generative AI struggling to produce an output. Therefore, queries that are too general or lack direction will result in poor outcomes. Similarly, a too-narrow, or overly specific prompt will produce equally unsatisfactory results, as this hinders the AI model’s flexibility.

An example of a bad prompt that is too general would be something like “Write a social media post for me.” In this example, there is no context, nor is there a sufficient foundation on which to build.

Instead, a better prompt would be something along the lines of “You are a [add a job title]. Draft an engaging LinkedIn post about [add a topic]. The target audience for this post is [add a target audience]. Include three high-performing hashtags. Here’s an example: [add an example].”

• Provide sufficient context (more on this below)
• Provide sufficient constraints, such as the approximate length you want the content. A simple sentence such as “The output should be about 100 words” will suffice.
• Do not provide more than a few constraints or you will limit the AI’s ability to construct something useful 

Context clues

Adding contextual information will only help a generative AI model render more suitable responses. Consider including the following:

• Give the prompt a persona. Tell it “you are a sports statistician,” or “you are a medical writer.”
• Is it intended for general readers or specialists such as doctors or engineers? Define your target audience.
• Communicate your purpose for creating the content. Will it inform, persuade, or debate a certain topic?
• What tone of voice should the content use? Exciting? Friendly? Let the AI know.
• AI is capable of paraphrasing very well, so paste an example to be mimicked into the prompt itself and mention that you want the output to be similar.

Avoid unethical queries

This is a no-brainer, but is still worth reinforcing. Not all generative AI models can perform a cost-benefit analysis or identify the possible harmful impacts of their responses.

• Avoid asking illegal, unethical, or dangerous topics

Monitor responses

Even with the best prompt writing practices, generative AI models are still liable to go adrift, creating off-track responses. Therefore, it’s best to monitor and review the output and refine it appropriately as needed. Keep an eye on the following:

• Is the output what you expected it to be? If it falls short, you likely need to add more context, so expand and clarify your prompt.
• Were false claims or inflammatory statements generated? If so, identify the part of the prompt that might be causing this.
  
• Was information repeated over and over? Ask for the output to be condensed or chunk the prompt into different sections.

Where does Narrativa come in?

Narrativa is a leading generative AI company with a cutting-edge platform that comes loaded with many features that help you analyze large volumes of data, generate reports, write in different tones, and more. The company has established its presence with tools and solutions servicing several industries, like life sciences, finance, marketing, and entertainment. Also, it has partnered with and provided services to many of the leading organizations and companies across the aforementioned areas, like Microsoft, the Wall Street Journal, the Leukemia & Lymphoma Society, and many others.

In addition to the solutions listed on our website, our clients approach us to help them automate many repetitive tasks so their teams’ attention can be directed toward higher-value projects.