How to Get Better AI Answers When You Don't Know the Right Question
How to Use AI to Find the Right Question Before You Ask
Most people treat AI like a search engine with better grammar. They type what they want, roughly, and hope the model fills in the gaps. Sometimes it does. More often, it produces something technically correct and completely unhelpful, because the question was too vague to answer well.
The fix isn’t learning to write better prompts. It’s learning to use AI to find the right question before you ask for the answer. The model is better at surfacing what you’re actually trying to solve than most people are at describing it cold.
These prompts run in sequence. By the end, you’ll have a question specific enough to get a genuinely useful answer, and a method you can repeat whenever you’re stuck at the fuzzy front end of a problem.
Step 1: Get the vague thing out of your head
Prompt 1: Problem dump processor
What it does: Takes an unformed, half-baked description of what you’re trying to figure out and turns it into a set of more specific sub-questions you might actually mean.
When to use it: When you know something is wrong or something needs solving but you can’t articulate what, exactly. The feeling of “I need to figure out X” with no clear sense of what X is.
The prompt:
I’m trying to figure something out but I can’t articulate it clearly yet. Here’s my messy version: [WRITE YOUR VAGUE PROBLEM IN PLAIN LANGUAGE, AS MESSY AS IT IS]. Don’t answer this yet. Instead, identify 5-7 more specific questions I might actually be trying to answer. For each one, write the question in one sentence and note in brackets whether it’s primarily a knowledge question, a decision question, or a diagnosis question.
How to use it:
Write your vague problem as you’d describe it to a friend who’d be patient with you. Don’t clean it up. Messy is better here
Paste it in place of [WRITE YOUR VAGUE PROBLEM IN PLAIN LANGUAGE, AS MESSY AS IT IS]
Read through the 5-7 questions and mark the one or two that make you think “yes, that’s closer to what I mean”
Example input: “I’m not sure if my newsletter is working. Like I’m writing it and people are reading it but I don’t know if it’s actually doing what I want it to do or if I should change something or if I’m just not being patient enough.”
What you’ll get: Five to seven specific questions pulled from the vague one. Something like: “Is your newsletter growing at a rate consistent with your goals? (decision)” and “Do you have a clear definition of what ‘working’ means for this newsletter? (diagnosis)” and “Are the right people subscribing, or just a lot of people? (diagnosis).” One of those will land closer than the others.
Advanced note: The question type in brackets matters. Knowledge questions (”what is X”) and decision questions (”should I do X”) need completely different prompts. If you try to answer a decision question with a knowledge answer, you’ll get information that doesn’t help you move. Sorting by type before you go further saves time.
That prompt does the thing most people try to do in their head and can’t: it breaks a vague feeling into specific questions you can actually work with.
The rest of the chain takes those questions further.
Prompt 2 — Assumption surfacer: Identifies every assumption baked into your chosen question, because a question with a false assumption in it produces a confident answer to the wrong thing.
Prompt 3 — Question sharpener: Adds the context, constraints, and decision criteria your question left out, then rewrites it as a fully-briefed prompt specific enough to get a genuinely useful answer.
Prompt 4 — Unknown-unknowns prompt: Surfaces the 3–4 things most people in your situation don’t know to ask about, that would significantly change the answer if they did.
Prompt 5 — Decision reframer: For decisions you’ve been stuck on, finds out whether the stuckness is about the options or about how the decision is framed, and whether you’re choosing between the real options at all.
Prompt 6 — Frame-breaker: Challenges the premise of your question entirely, for situations where the problem isn’t how you’re asking but what you’re asking about.
Plus: the question-building template you copy once and use on any problem where you’re not sure what you’re actually asking.