Most prompts fail because they are missing one of these: Role, Task, Context, Reasoning, Output format, Stop conditions. Add all six (with clear section headers) and ChatGPT 5.2 stops guessing and starts executing.
Role You are [specific role]. Your goal is [what success looks like]. Voice and style: [tone, reading level, format preferences]. Task Do this: [one sentence task]. Deliverable: [what you will produce]. Context Background: [2 to 6 bullets]. Audience: [who this is for]. Constraints: [time, budget, tools, policies]. Must include: [non-negotiables]. Must avoid: [landmines]. Reasoning Use this approach: State assumptions if needed Consider at least [2 to 3] options Explain tradeoffs briefly Check for obvious errors or missing steps Output format Return: [bullets/table/steps/script]. Include sections: [A, B, C]. Make it skimmable. Stop conditions Stop when: [definition of done]. If blocked: ask me [N] clarifying questions.
Read my Reddit post on this topic with pro tips and top use cases https://www.reddit.com/r/promptingmagic/comments/1pwsrmq/this_is_the_absolute_best_way_to_prompt_the/
Co-founder of Prompt Magic and ThinkingDeeply.ai Career Chief Marketing Officer
Build your personal prompt library, save your favorites, and access curated AI prompts created by the community
Access a vast library of high-quality AI prompts for every use case
Save prompts to your personal library and organize them your way
Get started with full access to our core features at no cost
No credit card required • Free forever • Join 10,000+ users
Use this prompt template for ChatGPT 5 to get the best results.
ChatGPT-5.1 is different. It reasons deeper, takes initiative, and can behave like a true teammate — but only if you prompt it correctly. This template forces the model to think like an expert, gather missing context, make assumptions, avoid filler, and produce complete, useful outputs instead of half-baked advice.
Paste this and fill the brackets:
The master prompt template (use this for top 1% results)
Use this template to get the best results from Claude Fable 5
When you force structure before the answer, quality jumps. Same model, same context - totally different depth.