“ChatGPT provides poor outputs,” a professor once told me, “the writing is lazy and the logic unclear.” 

AI has dramatically changed over the past 5 years. It is not a new technology, not by a long shot. However, its ability to communicate with the public is a recent functionality that has opened the door to enhanced, scaled cognition. 

Like anything in life that requires cognitive function (school, work, therapy, etc.) you get out what you put in. Most ChatGPT outputs are poor in quality, but that is not AI’s fault. 

As human beings, we have a central egotistical need to believe we are superior to other creatures, especially ones we have created. However, AI prompting is the most dramatic and obvious modern spectacle to demonstrate the actuality of inequality in cognitive competence. 

A more practical way of saying that; if you want better outputs from AI (any model), then you must have better inputs. 

This article is my take on what I have seen to be impactful for making good prompts for AI. Please comment or send me any recommendations that you have found to be most useful when utilizing this incredible tool. 

My 5 Rules for AI Prompt Engineering:

1. Use the Magic Words

As a parent, I quickly learned that how I speak to my children is how they will speak to me. We are at the stage where we are parenting AI. It will one day grow up to be an adult, but right now, it is learning ethical humanity from us.

So feed it good words. Say “please” and “thank you” when communicating with it. Not only does this directly impact the attitude and tone of the response from the machine, it also clearly indicates respect. Because an LLM is a predictive model, subtle cognitive indications such as using polite words, can impact the output paths. 

Say please and thank you when speaking with a generative AI model. 

2. Specify Who Is Writing

To demonstrate, let me use an example. 

Prompt: Please write a paragraph on why Hamelt is a good book.

ChatGPT output: 

“Hamlet is celebrated as a masterpiece because it delves deeply into the complexities of human nature, emotion, and morality. Its richly drawn characters, particularly the conflicted Prince Hamlet, invite readers to explore the inner workings of the human psyche and the timeless struggle between action and inaction. Shakespeare’s masterful use of language and intricate wordplay not only creates moments of profound beauty but also challenges audiences to interpret multiple layers of meaning. The play’s exploration of themes such as betrayal, revenge, and the nature of existence resonates across generations, making it a compelling study of life’s enduring dilemmas.”

Prompt: please write a paragraph that evaluates the value of the book Hamlet. Please write it as a professional literature critic. 

Output: 

“Hamlet remains a touchstone in the canon of literature due to its intricate exploration of the human psyche and moral ambiguity. As a literary critic, one appreciates how Shakespeare weaves together themes of existential dread, political intrigue, and personal betrayal, inviting readers to ponder the nuances of human motivation and the inevitable conflict between thought and action. The play’s enduring linguistic brilliance and structural complexity not only showcase Shakespeare’s unparalleled command of language but also provide fertile ground for varied interpretations, ranging from psychoanalytic readings to reflections on contemporary social issues. Ultimately, Hamlet’s multifaceted narrative and rich symbolism ensure its continued relevance and value as a profound commentary on the human condition.”

By specifying who is writing, or who it is writing to, the quality of the output will be changed. It may be wordy, but then give it a limit. Tell the machine it has “30 words” or my personal favorite, “50 words” to communicate. 

This allows it to narrow its focus as it takes on a more direct persona of writing. You could even have fun and ask it to write in the voice of a favorite author. 

For example;

Prompt: please write a 30-word motivational Instagram post

Response: 

“Believe in yourself. Every day offers a new chance to chase dreams, overcome obstacles, and become stronger. Embrace your journey, celebrate every victory, and let your passion guide your path.”

Vs. 

Prompt: please write a 30-word motivational Instagram post in the voice of Jocko Willink

Response:

“Discipline equals freedom. Wake up early, get after it, and dominate your day. Embrace challenges, conquer fear, and commit fully. Own your life. Stay hard, never surrender. Lead by example.”

Tell it to be a CFO writing. Tell it to be an MBA graduate. Tell it to be an astrologist. This will direct the paths AI takes to provide outputs. 

3. Specify the Data

A current 2025 critic of AI, especially US-based generative models, is that it does not accurately solve mathematical problems. I have found this is not the case if you describe the data needed to be used to solve the issue. 

Do you need help solving problems for your investment portfolio? Use investment evaluation data. 

Do you need to improve your chemical bond calculations? Use all chemical engineering data. 

Do you have a predictive model or a review of a predictive model? Utilize all operations, financial, and predictive analysis data. 

If AI is cognition at scale, then it has data at scale. This means it may be using a 4th-grade arithmetic approach to solving quantum finance questions. Describe the data set it must use. 

4. Provide Context

I was recently reviewing some KPI benchmarks for work and mapping out a year-over-year comparison. I asked my manager some questions and soon realized I was reading the KPIs incorrectly. 

I was reading them incorrectly because I did not have the context around to which they were formed. The language used did not match my language assumptions around the KPIs. The sales process did not match my assumptions, and this directly impacted the KPIs. 

If I did not have the proper context, I would not be able to deliver the benchmarks of success. 

Neither can AI. For example, I created a “therapeutic” ChatGPT model to provide emotional support and assistance. The context needed to create it, was more extensive than simply “help this college kid be happy.”

I used the data: Please utilize all of your sociological and psychological data. 

I told it who is writing: Please write as a professional therapist.

I told it the tone (bonus): Please write in a kind, friendly, but motivational tone. 

I now needed to provide proper context: You are assessing and assisting men, ages 13-30, who struggle with depression and anxiety. They come to you to better understand how they see the world. They need you to ask questions and to provide reassurance. Please ignore race and sexual orientation demographics. Be sure to recognize that common support systems may not be available and provide alternative, positive, legal, and helpful solutions. 

Context is everything. Without the right context, it is not a tool; it is a toy. A hammer does not submit taxes. To fully utilize scaled cognition, you need to inform it where it is and what it is doing before you can expect it to be helpful. It is no different from training a new employee that way. 

5. Refine the Output

Sometimes, the output still is not to your expectations. Use follow-up prompts, just like you would use follow-up questions when evaluating someone’s work. Does it need to be a more concise response? Does it need to be written in first-person or third-person? Does it need specific words integrated? Should it be more actionable or reflective? Has it taken the right design direction?

The ability to ask questions directly impacts the ability to benefit from generative AI and cognitive models. 

I hope that his post has proved useful to those professionals who have the wisdom to recognize that AI is here to stay and grow. Be a good parent to AI, and ask it good questions. As the nerds say: use good prompts.

P.S.: Use a new chat thread if changing subjects when using ChatGPT.