AI · prompts · productivity
The one prompt that turns AI into a thinking partner (not a vending machine)
AI isn't overhyped — you're just using it like a search bar. Here's the exact prompt that makes it interview you first.

okay, real talk: if AI has ever felt overhyped to you, I promise it's not the AI. It's the way we were trained to talk to it. Google taught us one move for years — ask a question, get an answer, close the tab. So we walk up to the smartest tool we've ever had and treat it exactly like a search bar. One prompt, one response, judge it, leave.
but AI isn't a search engine — it's a thinking partner. All the magic lives in the back-and-forth we keep skipping. So here's the one move that genuinely changed how the tool works for me (and for a lot of people I've shown it to): before it answers anything, you make it interview you first. Paste this at the start of your next chat and watch it get sharp fast.
The prompt
Paste this as your first message Before you answer, I want you to act as a sharp thinking partner, not a search engine. Do NOT give me a solution yet. Here's what I'm trying to do: [describe your goal in one or two sentences — e.g. "write a caption for a Reel about AI" or "plan a birthday weekend on a budget"]. First, ask me questions one or a few at a time until you actually understand what I'm trying to accomplish, who it's for, what a great outcome looks like, and any constraints (time, budget, tone, format). Keep asking until you're confident you have enough context to give me a genuinely excellent result. When you feel you have enough, say "I think I've got it" and briefly summarize your understanding back to me so I can confirm or correct it. Only after I confirm should you give me the final output. Assume I'd rather answer three good questions now than get a generic answer I have to redo. Let's go.
How to use it
Open a fresh chat in Claude or ChatGPT and paste the whole prompt in as your very first message.
Swap the one bracket — [describe your goal] — for whatever you actually want. One sentence is enough; the AI will dig out the rest.
Answer its questions honestly, even the ones that feel obvious. This is where the magic happens — you're feeding it the context it used to just guess at.
When it says "I think I've got it" and summarizes, read it. If it's off, correct it before you let it run. If it's right, tell it to go.
Keep the back-and-forth going after the first output too — 'make it warmer,' 'shorter,' 'more me.' That's the partnership, not a one-and-done.
that's it. That's the whole shift — you stop handing it a question and hoping it reads your mind, and you start letting it pull the context out of you first. Try it on the very next thing you were about to Google-brain your way through, and come tell me how it went. If this made something click, there's a lot more where it came from.