Every example below is something a real person actually used AI to do this week. Click to expand, see the weak prompt most people would write, then the strong one that actually works. Copy, adapt, and try it.
Why this works: Clear goal. You've told it what "good" looks like. You've told it what it can't know, so it'll flag uncertainty instead of bluffing. Verify one thing: pick one line item it flagged and call the billing number on the bill to confirm.
Why this works: The weak one gets you a generic template from the internet. The strong one gets you something that sounds like you. Verify one thing: read it out loud — does it sound like something you'd actually say? If not, edit.
Why this works: Real constraints get real answers. Verify one thing: check one recipe against what you know — is the protein portion realistic for 4 people? If yes, trust the rest.
Why this works: Turns AI into a patient lease-reader, not a lawyer. Verify one thing: if it flagged a clause as unusual, Google the clause + your state. Or, better — ask a friend who's rented in your state or your city's tenant rights hotline (most cities have one).
⚠️ Important: AI is not a lawyer. For a lease with stakes — long-term, high rent, or anything that feels off — talk to an actual attorney or a tenant-rights organization. Use AI to prepare the questions, not to make the decision.
Why this matters: AI will solve homework in 2 seconds. That's the wrong use. The right use is a private tutor that walks your kid through it, one step at a time, without judgment. See For Parents for the full framework.
Why this works: AI isn't a great fact-checker (it has a training cutoff, can hallucinate), but it IS a great thinking partner. This prompt uses it for the second thing. Verify one thing: take one factual claim and check it against a primary source.
Why this works: "Improve my resume" gets you 100 generic suggestions. This gets you 5 real ones, targeted. Verify one thing: the rewrites should still be true. If AI embellished anything, edit it back.
Why this works: AI is great at vocabulary and bad at diagnosis. This prompt uses it for what it's good at. Verify: don't. Take your questions to your doctor. That's the verification.
⚠️ Important: AI is not a doctor. For anything symptoms-related, urgent, or that involves actual treatment decisions — talk to your doctor. Use AI to prepare for the conversation, not to replace it.
Got a task that doesn't match any of the above? Use the prompt builder — type your rough task, answer the three questions, get back a prompt you can paste into ChatGPT (or Claude, Gemini, Copilot — whichever you prefer).