Effective Asking: Get Better Answers in Fewer Turns

Why asking well beats signing up well

Many people burn tokens on the same vague request. Effective asking means externalizing your success criteria so ChatGPT can aim at a real acceptance bar.

Thirty-second pre-flight

1. What is the deliverable (email, table, code, plan)?
2. Who is the reader or end user?
3. What hard constraints exist (length, tone, stack, compliance)?
4. What does “good” look like (example or scoring dimensions)?

Put those answers into the first prompt.

Five high-hit structures

Goal + material + format

Goal: Turn the weekly report below into a 120-word executive summary.
Material: ...
Format: one paragraph + three bullet takeaways.

Decision comparison

I must choose A vs B. Score cost, risk, time-to-ship, and maintenance (1–5). Recommend with stated assumptions.

Diagnose + fix

Here are logs and code. (1) Most likely root cause, (2) minimal patch, (3) how to verify.

Role with sharp boundaries

You are a strict code reviewer. No pep talk. List defects by severity only.

Explicit uncertainty

Mark uncertain facts as “uncertain”. Do not invent statistics or legal citations.

Follow-up pipeline

Turn	Your job	Example
1	Structure	“Outline only, no prose yet”
2	Depth	“Expand section 2 with an example”
3	Style	“More formal; remove filler”
4	QA	“Self-check against these 5 criteria and revise”

Say what to keep and what to change—avoid full restarts every turn.

Quality checklist

• On-topic?
• Any unverifiable precise numbers?
• Paste-ready format?
• Actionable next steps?
• Constraints respected?

Ask the model to self-audit against the list before the final draft.

Low-efficiency habits

• Ten unrelated tasks in one message
• “Make it professional” without a standard
• No negative examples or preferred samples
• Treating outputs as ground truth for prices, policy, or medical/legal claims

For accounts and billing, stay on chatgpt.com. Alternative access notes may appear later as a mirror site entry.

Practice: weak vs strong

Weak: “Do a competitor analysis.”

Strong: “Compare ChatGPT, Claude, and Gemini for a Chinese content team writing weekly reports. Dimensions: long-form quality, multi-turn rewrite, table cleanup, price sensitivity (qualitative). Output a comparison table, best-fit users, and anti-fit scenarios. Do not invent market share.”

Takeaway

Effective asking = success criteria + iterative refinement + forced QA. Free quotas go further once you ask this way; stronger models then compound the gains. Read Prompt Engineering Basics next.