The 6-step AI prompt formula gets you a better answer by building the problem before you ask for the solution: your situation, what has already failed, your constraints, the outcome you want, the expert you want answering, and only then the request. The guide's examples were written in Claude, and none of the six steps depends on which assistant you use.
Below are the 6 steps, the fill-in prompt for each, and what each one stops the assistant from doing.
We publish AI systems like this for 300,000+ senior professionals at AI Central, and more of them live in the AI Central Library.
Why most prompts get generic answers
Most people ask an assistant for the answer in one line and give it nothing else. It has no idea what you are working on, what you have tried, or what you can afford, so it fills those gaps with the most common case. The result reads well and fits nobody in particular.
The usual fix is better wording. It rarely helps, because what is missing is context, not vocabulary. The formula below supplies that context in a fixed order, one short paragraph per step.
How to write a better AI prompt in 6 steps
Paste the six paragraphs into a single message, or send them one at a time if you are working the problem out as you go. Replace everything in square brackets.
1. Describe your situation
Start with where you are, so the assistant works from your real starting point instead of assuming one.
I'm currently [describe exactly what you're doing - project, task, or goal]. Right now, I'm struggling with [clearly explain what isn't working or where you're stuck]. The main issue is [be specific about the problem, results, or gap].
2. Extract lessons from failure
List what you have already tried and why it did not work. Without this, the first suggestions you get back are often the ones you have already ruled out.
I've tested [list specific methods, strategies, or tools]. These didn't work because [briefly explain why - low results, wrong audience, inconsistent performance, etc.]. I want to avoid suggestions similar to these.
State your budget, time, tools and anything that cannot change, so the answer is one you can carry out.
I'm working with [state your budget, time limit, tools, or resources]. I need to stay within [mention any strict limits or non-negotiables]. Any solution should fit these constraints and be realistic to execute.
4. Define the real outcome
Name the business goal behind the task. The assistant can only aim at the result you want if you say what it is, and why it matters.
The objective isn't just [surface-level task], but rather [clear, meaningful outcome]. I'm aiming to achieve [specific result or metric], because [briefly explain why this matters].
5. Set the persona
Tell the assistant which expert to answer as, and in what style. A senior growth strategist, a copywriter and a product marketer would each give you different advice on the same problem.
Act as a [specific role or expert, e.g., senior growth strategist, copywriting expert, product marketer]. Your approach should reflect [mention tone, experience level, or style - strategic, data-driven, creative, etc.].
6. Trigger the final solution
Only now ask for the answer. The assistant already has your situation, your dead ends, your limits, your goal and a point of view, so the request itself can be short.
Now give me your best solution: Based on everything above, provide a clear, actionable plan I can follow. Keep it [mention preference - step-by-step, concise, detailed, etc.] and focused on achieving the stated goal within my constraints.
What actually changes
Before: one line in, a generic answer out, and several follow-ups spent explaining what you could have said at the start.
After: the assistant has your starting point, your dead ends, your limits and your goal before it writes anything, so the first answer is built around your situation. Save the six paragraphs as a template and reuse them for the next problem.
For everyday requests, Chapter 5 of the AI Bootcamp covers the three parts every prompt needs: context, task and format. For longer pieces of work in Claude, the Ultimate Claude Prompt Framework adds references, rules and a plan check before anything gets written, and these Claude power codes cover the shortcuts. More systems like this one live in the AI Central Library.
Frequently Asked Questions
What is the 6-step AI prompt formula?
A fixed order for building a prompt: describe your situation, say what has already failed, share your constraints, define the real outcome, set the persona, then trigger the final solution. Each step is one short paragraph with blanks to fill in.
Does the prompt formula work in ChatGPT as well as Claude?
Yes. The steps change what you tell the assistant, not which assistant you use. The guide's examples were written in Claude, and the same six paragraphs can be pasted into ChatGPT or Gemini.
Why should I tell AI what I have already tried?
Because otherwise it suggests the obvious options first, and the obvious options are usually the ones you have already tested. Listing them, with the reason each failed, removes them from the answer.
Do I need all six steps for every prompt?
No. A quick factual question needs one line. Use the full formula for problems where a generic answer is no use to you, such as a plan, a strategy or a decision with real limits on budget and time.
