Build a perfect prompt in 30 seconds — works with ChatGPT, Claude, Gemini & more.
📝 Blog Post📧 Email🏪 Product Description💬 Social Post
⚡ The Golden Rule: every prompt needs all 4 elements — Role + Context + Task + Constraints. Missing any one produces generic output. For best results, add a sample of YOUR writing as a style reference.
Why four elements, and why each one matters
Most weak prompts are not badly worded — they are missing one of four elements. The generator above assembles all four because each one removes a different kind of failure:
Role sets the standards the output is judged against. "Act as a senior SEO content writer" produces different choices about structure and depth than no role at all.
Context supplies what the model cannot know — your audience, your situation, what happened before. Missing context is the single most common cause of generic output.
Task defines done. "Write about X" has no end state; "write a 120-word welcome email with one CTA" does.
Constraints do the actual work. Tone, length, banned words, format — these are what make output usable without a rewrite.
A worked example
Here is what the four elements produce together, versus the same request without them:
Without structure: "Write a product description for a coffee dripper."
You will get something serviceable and forgettable, heavy on words like premium and high-quality, with no specifics — because none were supplied.
With all four: "Act as a direct-response copywriter. Context: a product description for a ceramic pour-over dripper sold to home brewers who already own a grinder. Task: a 150-word description that leads with the benefit. Constraints: lead with the outcome, include two checkable specifications, handle the paper-taste objection, avoid the words premium and high-quality."
Same tool, completely different output — because the constraints tell it what to avoid and what to include.
Five habits that improve any prompt
Give a style sample. Paste 200 words of your own writing and say "match this". Adjectives like "professional but friendly" get interpreted as generic business prose.
State the length explicitly. "About 150 words" beats "brief", which the model reads loosely.
Name the words to avoid. Banning premium, high-quality, seamless and cutting-edge removes most of the filler in one line.
Ask for one section, not the whole piece. Long one-shot requests produce filler in the middle. Constrained sections stay dense.
Ask it to critique before it rewrites. "List the problems, do not rewrite yet" produces better final output than asking for a rewrite immediately.
What no prompt can fix
Facts. Statistics, citations, prices and specifications are generated plausibly, not retrieved. Anything checkable must be checked by you.
Experience. "We tested this for 30 days" cannot be produced by a model that has not tested anything. That part is yours.
Judgement. Whether an argument is right, whether a claim is fair, whether the piece is worth publishing — a model will confidently produce all three badly.
Use the generator above to get the structure right, then do the parts that require you. That split is where the time saving actually comes from.