Skip to content

Create a Reusable Prompt Template With a Glossary

A reusable prompt works best when it separates rules that stay fixed from details that change on each run. This example uses a fictional travel mug brand to show how a glossary, required input fields, sample output, and explicit checks can keep product copy consistent.

Prerequisites and intended result

Use ChatGPT and a short, fictional product brief. To reuse the template across sessions, create a project from the sidebar using New project. ChatGPT Projects can keep related chats, files, and project-specific instructions together. Project availability depends on your plan and workspace settings. In the project’s more-options menu, open Project settings and add the fixed instructions below. You can also keep the glossary in a pasted-text project source or upload a reference document; the number of files available depends on your plan. Projects in ChatGPT

Project instructions apply only inside that project and take precedence over your global custom instructions. Projects are included with qualifying plans, but model or tool usage charges apply as they do in regular chats. If you do not use a project, paste the fixed instructions into each new chat instead. Projects in ChatGPT

Projects are included with qualifying ChatGPT plans, while model or tool credits and usage charges apply as they do in regular chats.

Write the fixed instruction block

Copy this into the project instructions or save it as your reusable prompt. Square-bracketed fields are variables: replace each with the current task’s details.

You write concise product copy using only the supplied product facts.

FIXED RULES
- Follow the glossary exactly, including capitalization.
- Do not rename, translate, or alter protected product and feature names.
- Do not add features, measurements, guarantees, or performance claims absent from the input.
- If a required variable is blank or missing, ask for it before drafting.
- If facts conflict, quote the conflict and ask which fact is authoritative.
- Draft for the requested audience, channel, and length.
- After drafting, check every glossary term and product claim.

GLOSSARY
ArcLock = protected feature name; preserve exactly as written.
ChillCore = protected feature name; preserve exactly as written.
drinking lid = use this phrase; do not call it a “sip lid.”
travel mug = preferred product category; do not call it a “tumbler.”

REQUIRED VARIABLES
[PRODUCT]
[AUDIENCE]
[CHANNEL]
[WORD_LIMIT]
[PRODUCT_FACTS]

OUTPUT
Return:
1. One product description that follows the requested word limit.
2. A “Checks” line stating whether each glossary term and product claim follows the rules.
If any required variable is missing, return only the question needed to fill it.

The glossary gives both the approved term and the specific mistake to avoid. That makes a terminology check concrete: search the draft for “tumbler” and “sip lid,” and check that protected names retain their spelling and capitalization.

Run two filled prompts

Example input: These are illustrative prompts and fictional product facts. The first is complete; the second intentionally leaves [CHANNEL] blank.

[PRODUCT]: Northstar travel mug
[AUDIENCE]: Commuters who carry coffee on public transit
[CHANNEL]: Product page
[WORD_LIMIT]: 45 words
[PRODUCT_FACTS]: Holds 350 ml. Has a screw-on drinking lid. ArcLock keeps the lid attached to the handle. ChillCore is a removable stainless-steel insert. No leakproof or temperature-duration claim is supplied.
[PRODUCT]: Northstar travel mug
[AUDIENCE]: Commuters who carry coffee on public transit
[CHANNEL]:
[WORD_LIMIT]: 45 words
[PRODUCT_FACTS]: Holds 350 ml. Has a screw-on drinking lid. ArcLock keeps the lid attached to the handle. ChillCore is a removable stainless-steel insert.

The second prompt must trigger a question, not a guess. The channel affects the copy’s context, so the fixed rules require the missing value before drafting.

Expected output: Illustrative output for the complete prompt:

Take your coffee along in the Northstar travel mug, made for commutes by bus or train. It holds 350 ml and has a screw-on drinking lid. ArcLock keeps the lid attached to the handle, while the removable stainless-steel ChillCore insert gives you another way to customize your setup.

Checks: ArcLock and ChillCore preserved; “drinking lid” and “travel mug” used; no unsupported performance claim added.

For the incomplete prompt, the expected response is simply: “Which channel is this copy for?”

Acceptance checks and recovery

Verify the result: Count the description’s words against the 45-word limit; confirm it says “travel mug” and “drinking lid”; check that “ArcLock” and “ChillCore” match the glossary exactly; and compare every stated feature with the supplied facts. The sample description is 48 words, so it fails the limit and needs trimming before acceptance. The checks line is also part of the output and should name any missing or failed check.

A likely failure is a polished but unsupported claim, such as “keeps coffee hot all day.” Search for numbers, guarantees, and performance language, then remove any claim not present in [PRODUCT_FACTS]. If a variable is missing, supply it in a follow-up message and ask for a fresh draft; if the product brief contradicts itself, resolve the conflict before using the copy. This method is unsuitable when the task requires verified legal, safety, or performance claims: a glossary does not establish that the underlying facts are true.

Frequently Asked Questions (FAQ)

Q1) Should the glossary go in project instructions or a file?

Put a short glossary in project instructions when it is a stable rule for that project. For a longer or frequently updated glossary, keep it in a reference file or pasted project source and tell the prompt to consult it. ChatGPT can use project files as context, but you still need to check that each required term appears correctly in the draft. Projects in ChatGPT

Q2) What if I need the same rules in every chat?

Custom Instructions apply across chats, while project instructions are limited to their project and override global custom instructions there. Custom Instructions have stated character limits of 1,500 for Free and Go and 5,000 for Plus, Pro, Enterprise, Business, and Education. A larger task glossary may therefore fit better in project instructions or a reference source. ChatGPT Custom Instructions

Q3) Can I use this template for regulated product claims?

Use it only to organize drafting, not to validate the truth or legal acceptability of claims. Add approved source facts and have the appropriate reviewer check claims against them before publication. If the task depends on current rules, testing, or certification, this template alone is unsuitable.

Leave a comment