AI Central's 6-Step AI Prompt Formula replaces the one-line question with a six-part brief. You describe your situation, list what you already tried and why it failed, state your constraints, name the real outcome behind the task, assign the model an expert persona, then ask for the plan. Each step closes off a specific way an answer goes generic. The premise AI Central leads with is blunt, your context determines the output.
Reviewed August 2026.
The problem the formula is built for
Most prompts fail for a boring reason. The person writing knows the situation and the model does not, so the model fills the gap with the most average answer available. It suggests the tactic everyone suggests. It proposes a budget you do not have. It solves the task you named instead of the outcome you actually wanted.
AI Central's fix is not a phrasing trick. It is a sequence of six statements, each one closing a specific gap, ending in a single request for the plan. The promise in the title is better outputs every time, and the mechanism behind that promise is context supplied deliberately rather than assumed.
Steps one to three put reality on the record
The first half of AI Central's formula is purely descriptive. Nothing in it asks the model for anything at all.
Step one is your situation. The template has you state exactly what you are doing, the project, task or goal, then where you are stuck, then the specific problem, result or gap. AI Central's stated purpose is to give the model a clear starting point. Three clauses, and it no longer has to guess which of a hundred possible scenarios you are in.
Step two is the failure log. You list the methods, strategies or tools you have already tested, briefly explain why each one did not work, low results, wrong audience, inconsistent performance, then state that you want to avoid suggestions similar to these. AI Central gives the step one line of justification.
Prevent the AI from suggesting things you’ve already done
This is the step most people skip and the one that buys the most. Without it, a model will confidently hand back the four tactics you already burned three months on, because those tactics are the most common answer in the training data and nothing in your prompt says otherwise. With it, the obvious tier is ruled out before the model starts.
Step three is constraints. Budget, time limit, tools and resources, plus any hard non-negotiables, closing with the instruction that any solution should fit these constraints and be realistic to execute. AI Central frames the goal as keeping solutions realistic and achievable. A plan you cannot fund is not a plan, it is a wish with bullet points.
Steps four to six aim it
The second half changes register. Here AI Central is shaping how the model thinks, not just what it knows.
Step four separates the task from the outcome, and it is the most underrated move in the set. The template is explicit about the split, the objective is not the surface-level task, it is the meaningful outcome behind it, together with the specific result or metric you want and why that matters. Ask for a better headline and you get a better headline. Say the objective is qualified signups and the model is free to tell you the headline was never the problem.
Step five sets the persona. Act as a named role or expert, and AI Central's own examples are senior growth strategist, copywriting expert and product marketer, then specify the approach the answer should reflect, whether that is tone, experience level or style. The point is to make the advice come from a specialised perspective rather than a general one.
Step six is the only step that asks for anything. Now give me your best solution, based on everything above, provide a clear actionable plan I can follow, kept in whatever format you prefer and focused on the stated goal within the constraints. AI Central labels the function of that step plainly.
Trigger the final, high-quality strategy
Why the order does the work
The sequence is not decorative. Five statements with no request, then one request that refers back to all of them, has a specific property, by the time you ask, everything needed to reject a bad answer is already on the record.
Reorder it and it degrades. Put the persona first, as most prompt advice does, and you get an expert voice with no information behind it. Put the ask first and the model answers before it knows the constraints, then spends the rest of the conversation walking that answer back. AI Central places the persona at step five, immediately before the ask, where it colours the one output that matters instead of the whole setup.
The failure log and the constraints work by subtraction. Each one deletes a region of possible answers, and what survives is narrower, more specific, and much harder for the model to have produced without you. That is the whole trick. You are not asking better, you are leaving less room to be wrong.
Using it without wasting your own time
The formula is a brief, and a brief costs ten minutes. That is the honest trade, so spend it where it returns. Four rules make it practical.
- Send it as one message, not six. The steps are written to stack, which is why the final one says based on everything above.
- The brackets are the work. Every step is a template with placeholders, and a lazy fill produces exactly the generic answer the formula exists to prevent.
- Reuse the first three steps. Situation, failures and constraints change slowly, so save them once and swap only the outcome, the persona and the ask.
- Skip it for small jobs. Rewriting a paragraph does not need a six-part brief. A strategy you are going to run for a quarter does.
For a beginner, the highest-value step is three, constraints, because unusable advice is the most common first disappointment with AI. For anyone further along, it is step two, the failure log, since the more experience you have the more of the obvious answer set you have already exhausted.
Where it strains
Two limits are worth stating. AI Central's formula is built for consulting-shaped questions, strategy, growth, positioning, anything where a plan is the deliverable. It is heavy machinery for a quick edit or a factual lookup.
And there is a labelling slip worth flagging so nobody follows it literally. The sixth step repeats the fifth step's heading, set the persona, while its stated purpose and its prompt are both the closing request for a plan. The function of step six is the ask, not a second persona.
Do I send all six steps at once or one at a time?
One message. The steps are written to stack and the sixth refers explicitly to everything above it, which only exists if it is all in front of the model. You can send them as a conversation instead, it just invites the model to start answering before you have finished the brief.
How long should each step be?
Short. Every template AI Central provides runs two or three sentences, and step one is three clauses. This is a brief, not an essay. Precision inside the brackets beats volume every time.
Does this only work with Claude?
AI Central's worked examples run in Claude, with the final step shown on Sonnet 4.6. None of the six steps is a product-specific instruction though. They are six statements about your own situation, your failures, your limits, your goal, and the voice you want back, which is information any model can use.
What if I have not tried anything yet, so there is no failure log?
Then say so, and use the space for what you have ruled out and why. The step exists to stop the model handing back ground you have already covered. If you have covered none, naming the approaches you do not want, and the reason, does the same job of narrowing the field.