Take the vague topic you started with and rewrite it as a single sentence that contains a who, a what, and a constraint — then paste that sentence into the AI tool along with the raw material you already have.
That's the whole method. Here is a complete prompt you can copy, using a dental office example: "You are helping a dental practice write a reminder email. Audience: patients who missed a cleaning appointment in the last 90 days. Goal: get them to rebook. Constraint: under 120 words, no guilt language, one clear call to action.
Here is the last reminder email we sent, which got a low response rate: [paste it]. Rewrite it so it fixes whatever made people ignore it, and tell me what you changed and why." That prompt works because every vague word has been replaced with a decision.
"Patients" became "patients who missed a cleaning in the last 90 days." "Better email" became "under 120 words, no guilt language, one call to action." The tool no longer has to guess what you want, so it stops producing the bland, generic output that makes people say AI writing all sounds the same.
Step 1: find the question hiding in what you already have
The fastest source of a specific question is not your head. It's an artifact you already own: an old email, a support ticket, a meeting note, a product description, a Slack thread. Read it and ask, "what decision does this thing fail to make?"
A missed-appointment email that nobody opened is telling you the question is not "write a reminder email." The question is "why did this reminder get ignored, and what would make someone act?" That reframe is the entire skill.
A topic is a noun. A question is a decision. AI tools respond well to decisions and poorly to nouns.
Step 2: the before/after pair
Vague prompt: "Write an email about dental cleanings." The tool has no idea who it's writing to, what happened before, how long it should be, or what success looks like. You'll get 300 words of cheerful filler that mentions "oral health" four times.
Specific prompt: the copy-paste version above. Same tool, same day, wildly different output — because you moved the thinking from the tool's guesswork into your sentence. A useful rule: if your prompt could apply to any company in the world, it's too vague. If it could only apply to your situation, it's specific enough.
Step 3: give it the raw material, not just the instruction
Most beginners write prompts as commands. Better prompts are commands plus evidence. Paste the old email, the customer complaint, the three bullet points from your notes.
This matters because language models complete patterns, and the pattern you supply in your prompt heavily shapes what comes back. If you paste a stiff, formal reminder, you'll get a stiff, formal rewrite. If you paste a warm one and say "keep this tone, fix the structure," you've given the tool a target instead of a blank page.
According to our AI tool database, which tracks 360 AI tools with a pricing and capability snapshot recorded at verification time, capability differences between tools are real — but the gap between a vague prompt and a specific one is usually larger than the gap between two tools.
When this method fails
It fails when you don't have an artifact to mine. If you're writing about something genuinely new — a product that doesn't exist yet, a policy with no precedent — there's no old email to paste, and you'll have to build the specificity from scratch by interviewing yourself: who is this for, what do they believe now, what should they believe after reading?
It also fails when the constraint is fake. Writing "under 120 words" when you actually need 400 produces a cramped, useless draft. And it costs time up front — five minutes of writing a real question versus thirty seconds of typing a topic.
The trade is worth it for anything you'll publish, and not worth it for a throwaway internal note. One more limit: a specific prompt improves the shape of the output, not its accuracy. If the tool invents a statistic about dental health, your well-written prompt didn't prevent that.
For that problem, the fix is a separate habit — How do I stop ChatGPT from making up fake facts in my content? covers it well. A zero-prompt AI content generator like AI-Mind skips this step entirely by generating from a topic directly, which is a different trade: less setup, less control over the specifics.