An AI prompt is the instruction you type into an AI tool to get a response — and a good first prompt has five parts: role, task, context, format, and tone.
If you only remember one thing, remember this: the AI has no idea who you are or what you want unless you tell it, so a prompt is less like a search query and more like a work order you'd hand to a new freelancer on their first day. You would never tell a freelancer "write something about my business" and expect gold. The same rule applies here.
Here's the formula spelled out. Role tells the AI what hat to wear ("You are a copywriter for a small bakery"). Task is the single thing you want done ("write a 100-word Instagram caption").
Context is the background only you know ("we sell sourdough on Saturdays and our customers are local families"). Format is the shape of the output ("a caption plus three hashtags"). Tone is how it should sound ("warm and a little playful").
Put together, that's one paragraph, and it takes about thirty seconds to type. Compare that to "write something about my bakery" — same tool, same model, wildly different result. The reason the formula works is that AI models generate text by predicting what comes next based on everything in front of them.
Every missing detail is a gap the model fills with the most statistically average guess. Role, task, context, format, and tone are simply the five gaps that matter most for everyday writing tasks.
Let's run the bakery example end to end. Your first prompt: "You are a copywriter for a family bakery. Write a 100-word Instagram caption about our Saturday sourdough drop.
Context: we sell out by 10am, regulars line up early, and we're in a small town. Format: caption plus three hashtags. Tone: warm, a little playful, no exclamation marks."
The output will mention sourdough, Saturday, and the sell-out — because you told it to. Now imagine the vague version: "write something about my bakery." You'll get generic copy about fresh bread and morning coffee that could belong to any bakery on earth.
Same tool. The only difference is the five parts. A useful tip most beginners miss: put the constraint after the task, not before.
"Write a 100-word caption... no exclamation marks" reads more naturally to the model than burying the length rule in the middle of a sentence, and you'll get fewer surprises.
Your first draft is almost never your final one, and that's normal — iterating is part of writing a prompt, not a sign you did it wrong. Three fixes cover most problems. If the output is too long, don't rewrite the whole prompt; just reply "cut that to 60 words and keep the sourdough detail."
If it's too vague, the missing piece is almost always context — add one specific fact, like "mention that we mill our own flour." If it's off-topic, your task was probably doing two jobs at once; split it into two prompts, one for the caption and one for the hashtags. This back-and-forth is the actual skill.
Beginners who treat the first output as final get mediocre results and blame the tool; beginners who treat it as a rough draft get usable work in two or three rounds.
Where this advice breaks down: the formula works well for writing, summarizing, and brainstorming, but it is weaker for tasks with a single correct answer, like math or factual lookups, where extra context can actually confuse the model. It also costs you time up front — a five-part prompt takes longer to write than a lazy one, and for a throwaway question that trade isn't worth it.
And no prompt formula fixes a tool that's wrong for the job. If you need an image, a text prompt won't help; according to the AI-Mind AI Tool Database, Midjourney is built for image generation, while Adobe Photoshop's Firefly tools handle AI-assisted photo editing — matching the tool to the job comes before matching the prompt to the tool.
The database tracks 360 AI tools, which is a useful reminder that "the AI" isn't one thing. Start with the five-part formula, iterate two or three times, and you'll be writing prompts that work within a week.