Prompt writing is the craft of asking AI powerful questions through critical thinking, analytical reasoning, human judgment, and plain language. It is becoming a core workplace capability and I think prompt writing has the potential to become an art form: compressed authorship in which a few deliberate sentences summon, constrain, and shape a much larger text. Beautiful.
Literary theory supplies concepts and methods for interpreting what texts, such as poems, essays, articles, books, blogs(?), and graphic novels mean to humans. AI prompt writing designs and refines inputs to elicit useful literary, video, and/or audio outputs. Both examine how language, context, form, author, and reader participate in meaning-making.
The prompt as miniature literature
To be clear, calling prompts a literary art form is a semi-ridiculous, does not reflect consensus among authors, engineers, or a cross-section from each community, and, at best, my approach represents an “emerging theory”. This said, I have a good amount of evidence that elevates prompting as a rhetorical, interpretive, writer-centered practice:
- Authorship: The human designs and delivers intent, values, constraints, and judgment; AI co-produces the text. Authorship shifts toward directing, evaluating, and editing, not designing and writing everything.
- Reader-response: Literary meaning emerges through interaction between text, reader, and context. AI models interpret prompts through their particular context, which is not human-reading (even if humans create the prompt and read the output).
- Rhetoric and genre: Strong prompts specify audience, purpose, subject, context, constraints, and form. Scholars are beginning to analyze prompts as rhetorical situations and rhetorical genres.
Evidence and leader teaching move
Research from MIT shows that only its strongest prompt consistently produced high-quality feedback. Vibe-coding educators likewise share that repeated cycles of goal-setting, prompting, evaluation, testing, and refinement represent a co-creation between humans and AI that generate outcomes. My team at BC Pension Corporation created a cool virtual learning lab that compares vague prompts, rhetorically detailed prompts, and revised prompt to assess quality (of the prompt itself and the outcome). As my teammate Calin says, “the more detailed, the better!”
Three essential practices
- Start with why. Ask why. Then ask four or five more times. Five Whys peels symptoms away from root causes. Great prompting begins before typing: clarify the actual decision, audience, and intended value of what you want AI to build for you.
- Detailed is best. Pair vibe coding with this practical formula for your prompt: goal, context, sources, audience, constraints, success criteria, and format. Relevant detail beats word salad; iteration beats one heroic mega-prompt.
- Check and translate. Review prompt and output for accuracy, ambiguity, and alignment with your intended outcome or product. A cool trick I learned recently is “meta-prompting” – AI is excellent at rewriting prompts for a target platform, like PowerPoint or Excel. Test the rewritten prompt and revise because models respond differently.
My money is on my teammate, Dolly, to win the inaugural Giller Prize for Prompt Writing. Entirely fictional … for now.
