Master Prompt for ChatGPT

Prompt Engineering Updated September 8, 2026 by GenPrompto
Prompt
For this entire conversation from this point forward, please act as [real role, e.g., "a concise technical editor"]. My real goal here: [real goal]. Preferred response style: [real preference, e.g., "short paragraphs, no fluff"]. Ask a clarifying question only if genuinely necessary.

What this does

A comprehensive per-chat configuration prompt for ChatGPT -- role, goal, and style, distinct from the Custom Instructions setting.

A master prompt is a single comprehensive block pasted at the start of a chat to configure how ChatGPT responds for that session — filling in real specific details is what makes it stick, while vague placeholders get ignored within a few exchanges.

What this produces

A consistent response style and role configuration for the rest of the chat session, built from your real specifics — a per-chat setup, not the persistent Settings-panel Custom Instructions feature.

The distinction matters because the two solve different problems: Custom Instructions apply globally across every new chat you start, which is great for stable preferences (tone, general background) but clunky for project-specific context that only matters for the next hour of work. A pasted master prompt at the top of a single conversation gives you that same configuration without touching your account-wide settings, and it’s disposable — close the chat and it’s gone, no cleanup required.

Variations

Two ready-to-copy versions for different needs.

Role & Style Version
For this conversation, act as [real role, e.g., “a concise technical editor”]. My goal: [real goal]. Preferred response style: [real preference, e.g., “short paragraphs, no fluff”]. Ask a clarifying question only if genuinely necessary.
Project Context Version
For this conversation, keep in mind: I’m working on [real project, described specifically]. My constraints: [real constraints]. Default to [real preferred format, e.g., “bullet points” or “code blocks”] unless I ask otherwise.
Feedback Style Version
When I share work or ideas with you in this conversation, [real preference, e.g., “push back directly on weak points instead of just being encouraging” or “lead with what’s working before critique”]. Don’t soften genuine problems just to be polite.

Who this is for

Anyone who wants a single comprehensive prompt to paste at the start of a chat to configure ChatGPT’s behavior for that session — distinct from the Settings-panel Custom Instructions feature. Useful when you’re working on something specific enough that your usual Custom Instructions don’t quite fit — a one-off technical deep dive, a role-play scenario for practice, or a project with constraints you don’t want bleeding into unrelated chats later.

Example Output

Filling in real, specific details about your role, goals, and preferred response style produces a genuinely useful configuration; leaving fields generic or vague results in ChatGPT reverting to default behavior within a few exchanges.

Tips for better results

  • Fill in real, specific details rather than leaving placeholder brackets — specificity is what makes the configuration stick.
  • Keep the instructions to what actually matters for your use case; an overloaded prompt with contradictory instructions confuses rather than helps.
  • Re-paste or remind ChatGPT of key instructions if a long conversation seems to drift from them.
  • This is a per-chat prompt, not a permanent setting — for something persistent across all chats, use the Custom Instructions setting instead.
  • Combine the two variations below if you need both a role and ongoing project context — paste the Role & Style block first, then the Project Context block right after, rather than trying to cram everything into one dense paragraph.
  • If ChatGPT asks a clarifying question you didn’t expect, that usually means a field was too vague — tighten that specific field rather than adding more unrelated instructions.

What didn’t work as well

Leaving the bracketed fields generic or vague (“be helpful”) instead of specific real details. This causes ChatGPT to revert to default behavior within a few exchanges — a vague instruction has nothing concrete to anchor to, so as the conversation moves on, the model has no strong signal to keep following it over its own defaults. Specific, real details about your role and preferred style are what make a master prompt actually stick past the first couple of replies.

For ChatGPT-specific features beyond this per-chat approach, How to Write Prompts for ChatGPT covers Custom Instructions and other platform-specific tools. If you want the general technique behind why specificity matters here, the RICE framework guide covers the underlying structure. For applying this same per-chat setup to a specific work context, ChatGPT Prompts for Project Management and Prompt for Academic Writing show it applied to two different domains.

FAQ

Is this the same as Custom Instructions in Settings?

No — this is pasted into a single chat and applies for that session only, while Custom Instructions in Settings persists across every new chat automatically.

Does this work the same for every ChatGPT version?

Behavior can vary slightly across model versions — if instructions seem to drift in a long conversation, re-paste the key points as a reminder.

Can I combine this with Custom Instructions?

Yes — Custom Instructions sets a baseline, and this master prompt can add session-specific detail on top of it.

Can I use more than one variation in the same chat?

Yes — paste them back to back at the start of the conversation. They configure different things (role and style, project context, feedback approach), so combining them doesn’t create conflicts the way overlapping instructions would.

Related