It can be intimidating opening ChatGPT, Perplexity, or other generative AI programs and seeing a blank screen.

Filling that blank screen only takes typing a simple prompt, like, “Tell me how to write a good email.”  Now, a limitation of generative AI is that they don’t come with instructions to guarantee the boundaries of what generative AI doesn’t make. To do that, you only need a specific prompt to put the boundaries on what you don’t want it to make.

This prompt is called a boundary prompt.

Boundary prompts are simple, intentional prompts. What I love about them is how versatile they are. Why do you need to have a boundary prompt? Simple. Setting clear boundaries for the generative AI to follow reduces its inaccuracies. Having those set boundaries increases your trust in the generative AI, bringing it a step closer to being a true assistant. Fortunately, you don’t need to code to make these improvements happen!

This week’s Brief breaks down using boundary prompts.

Putting Boundary Prompts Into Practice

As you saw in the video, building boundaries with a boundary prompt isn’t difficult.

By adding a boundary prompt within the prompt, or persona, you establish how you want the generative AI to work.

Let’s look at two examples of boundary prompts you can use today.

The first boundary prompt is a universal boundary for everyone that sets the boundaries you want across all of the generative AI’s work: “Do not lie, fabricate your answer, or hallucinate. If you cannot answer this, then type out, ‘This requires manual review,’ and continue your response.”

The second example is for company or client work. Use it to keep a boundary on what generative AI can’t make: “You’re not an authority on legal, ethical HR, or regulatory policy content. You will also cross-check against our company brand standards, compliance, and accessibility guidelines.”

As you can see, boundary prompts are a single step towards dramatically improving the quality of what generative AI does and doesn’t create. Boundary prompts allow you to “see” what the AI won’t make, reducing mistakes and hallucinations, which allows you to trust more of its content.

Now, the next time you look at the blank screen of ChatGPT or Perplexity, you’ll set your boundary prompts in place, making generative AI your assistant.