How to Write a Reusable Prompt: A 5-Part Template

Have you ever written the perfect prompt for ChatGPT or Claude, got exactly the result you wanted, and then lost the exact wording a week later when you needed it again? It's the most common problem for anyone who uses AI every day: you work prompt by prompt, instead of building tools you can reuse.

The solution isn't memorising your best formulas, but writing prompts from the start according to a repeatable structure. Just 5 fixed parts, always in the same order, are enough to turn a one-off request into a template that works every time.

Why a "disposable" prompt wastes your time

When you write a prompt thinking only about the immediate result, you tend to cram everything together: the context, the request, the desired tone, all mixed into a single paragraph. It works, but only that once: next week, with a slightly different variant of the task, you'll have to rebuild it from scratch. A template with separate, recognisable parts, on the other hand, adapts by changing only the pieces that need to change.

The 5 parts of a reusable template

A prompt designed to be reused is always built by separating these five elements, even just with a line break between one and the next:

• Role — who the AI should "be" while answering (e.g. an accountant, a copywriter, a contract lawyer).
• Context — the background information needed to understand the situation (the industry, the recipient, the starting constraints).
• Task — the specific action requested, described directly and verifiably.
• Output format — how you want the answer structured (a bulleted list, a table, a ready-to-send email, a maximum length).
• Example or constraint — a concrete example of the desired result, or a limit to respect (what to avoid, a tone not to use).

A practical example, from generic to reusable

A prompt written "on the fly" might be: write an email to chase a payment. It works, but it produces a different result every time and isn't easy to adapt quickly.

The 5-part version, on the other hand, keeps the same structure even when you change client or amount: Role (you are an administrator who handles payment reminders), Context (invoice number X, overdue by Y days, client with whom the relationship is good), Task (write a payment reminder), Format (ready-to-send email, maximum 120 words), Example/constraint (friendly but firm tone, no threat of legal action). Only the details in brackets change: the structure stays identical, and the result is predictable every time.

Where to save your templates

A template that only lives in your chat history is already half lost: next time you need it, you probably won't find it quickly. It's worth collecting them in a single document or spreadsheet, organised by use (emails, contracts, reports), so you can copy and adapt them in seconds instead of rewriting them from scratch. We covered how to build this collection step by step in our article on how to build a prompt library for your practice.

If you'd rather not start from scratch, our free prompt generator automatically builds the 5-part structure from a few pieces of information you enter, and it's useful even just to understand how to break down a request you already use. If you generate many different prompts every week, it may also be worth understanding how our token-based prompt generator works.

Frequently asked questions

Do you need to be good at writing to use this method?

No: the 5-part structure works precisely because it takes the weight off the writing itself. You don't have to compose elegant text, just fill in five simple, direct pieces of information.

Does the template need to be rewritten for each different AI (ChatGPT, Claude, Gemini)?

No, the 5-part structure works the same way with any conversational AI assistant: at most the ideal prompt length changes, not the logic behind how it's built.

How many templates is it worth having ready?

It's better to start with just a few templates for the tasks you repeat most often during the week, and add new ones only once you notice you're rewriting the same request for the third time.