# ChatGPT Custom Instructions vs Project Instructions for Work > Custom instructions apply to every ChatGPT chat; project instructions apply to one client's chats and win. Where each rule goes, and memory's role. Source: https://learn.ynteractive.com/content/blog/chatgpt-custom-instructions-vs-project-instructions · Updated 2026-10-17 · Free from Stepthrough (https://learn.ynteractive.com) Quick answer **Short answer:** put rules about *you* in [custom instructions](https://learn.ynteractive.com/content/glossary/custom-instructions) and rules about *one client or one job* in that Project's instructions. Custom instructions, in **Settings** under **Personalization**, apply to every chat you start. [Project instructions](https://learn.ynteractive.com/content/glossary/project-instructions) apply only to the chats inside that project, and they take priority over your general custom instructions. Memory is neither: it is what ChatGPT decides to save, so don't rely on it for rules. The free [ChatGPT Projects lesson](https://learn.ynteractive.com/content/chatgpt-projects-brand-voice) sets up project instructions in about nine minutes, in a practice copy. Last updated October 17, 2026 [Take the free module →](https://learn.ynteractive.com/courses/chatgpt-for-everyday-work/7)Module 7 of the free [ChatGPT for Everyday Work](https://learn.ynteractive.com/courses/chatgpt-for-everyday-work) course, about 9 minutes. No account needed. By [Robert Breen](https://learn.ynteractive.com/content/about) · Published October 17, 2026 · Updated October 17, 2026 · 5 min read ChatGPT · Projects · Settings The business problem is a slow leak. A bookkeeper who serves eight clients types the same background into chat after chat: who the client is, how formal to be, which price list is current. The day one line gets left out, a reply goes out with the wrong fee or the wrong tone. Putting each rule in the right place once is how you stop paying that cost every day. If you want the simpler question of how memory differs from custom instructions, see [ChatGPT memory vs custom instructions](https://learn.ynteractive.com/content/blog/chatgpt-memory-vs-custom-instructions). This post is about where your *work rules* should live. ## What is the difference between custom instructions and project instructions? Scope. Custom instructions apply to every chat, from a client email to a lunch recipe. In the [Turn a Great Prompt into Reusable Instructions lesson](https://learn.ynteractive.com/content/turn-chatgpt-prompt-into-project-instructions), that is exactly why they are ruled out for a job's rules: applying to every chat is too broad. Project instructions apply only to chats in that one project, which is just right for one client or one recurring job. | | Custom instructions | Project instructions | | --- | --- | --- | | Where | Settings, Personalization | Project page, the **⋯** next to Share, Project settings, Instructions | | Applies to | Every [new chat](https://learn.ynteractive.com/content/glossary/new-chat) you start | Only chats inside that project | | If they disagree | Lose | Win: they take priority over your custom instructions | | Good for | Your role, how you like answers | One client's tone, files to use, sign-off, rules like "never invent a price" | ## What should go in custom instructions? Things that are true about you in every chat. Your role, that you want plain answers, that you want a question back when a fact is missing. Nothing client-specific. A rule like "sign as Jess at Brightwater Cleaning" belongs to one business, so it does not go here. Custom instructions are personal. They shape your own chats, not your coworkers'. If the whole team needs the same rules for a client, that is another reason to put them in a project. ## What should go in project instructions? Everything that is true every time you work on that client or job. The lesson's instructions for a made-up cleaning company are a good model, five short lines: - who the business is - the voice: warm, plain, short sentences - use the prices in the price list file, and never invent a price or a discount - follow the brand voice guide - how to sign every message Write them like a note to a new assistant on their first day. Then add the files the instructions point to under the project's **Sources** tab, so every chat in the project can read them. ## How do you add project instructions? These are the steps taught in the [ChatGPT Projects lesson](https://learn.ynteractive.com/content/chatgpt-projects-brand-voice). Menus change, so look for the same names in your account. 1. Open the project. Click the **⋯** button next to **Share**. 2. Click **Project settings**. 3. Click the **Instructions** box and type or paste your rules. 4. Click **Save** at the bottom. Save and Cancel appear once you change something. Closing with the **X** would not save your instructions. When a result is off, change one rule in the instructions and every future chat in the project picks it up. Old chats do not change, so start a new chat inside the project to test it. ## Where does memory fit? Memory is what ChatGPT carries from one chat to the next, and ChatGPT decides what it keeps. It can be handy for background, but it can carry an old detail into a new task, so it is the wrong home for a rule you need followed every time. Two habits from the lessons help at work: - **Turn on project-only memory for client projects.** When you create a project, the **Default memory** setting (or a gear icon, on some accounts) lets you pick **Project-only memory**. Chats in that project then draw only on each other, and your other chats can't see them. - **Check saved memories now and then.** Memory lives under Settings, Personalization. The [Clear Client Data Out of ChatGPT lesson](https://learn.ynteractive.com/content/how-to-delete-chatgpt-history-clear-memory) shows how to open saved memories and delete one client's memory without using Delete all memories. On a shared work login, the lessons' advice is to turn **Enable memory** off. And remember that a Temporary Chat ignores memory and custom instructions, so it will not follow your standing rules either. ## Try it free, step by step **ChatGPT Projects: Keep Your Brand Voice, Files and Instructions in One Place**. Set up a ChatGPT Project with your brand voice instructions, your price list and your style guide, so every new chat already knows your business. You do every step yourself in a practice copy of ChatGPT, it checks your work as you go, and “Do it for me” finishes any step you get stuck on. Free, and nothing touches your real accounts. [Start the free lesson](https://learn.ynteractive.com/modules/chatgpt-projects-brand-voice) [Or read the full step-by-step guide](https://learn.ynteractive.com/content/chatgpt-projects-brand-voice) ## Key takeaways - Custom instructions are for you and apply to every chat; project instructions are for one client or job and apply only inside that project. - When they disagree, project instructions win. - Keep client names, prices and sign-offs out of custom instructions. - Don't rely on memory for rules; use project-only memory for client projects and check saved memories. - Edit the project's instructions once when a result is off, then test in a new chat. ## Frequently asked questions **Do project instructions override custom instructions?** Yes. In the lesson, project instructions apply to every chat in the project and take priority over your general custom instructions. **Where are ChatGPT custom instructions?** In **Settings**, under **Personalization**. They apply to every chat you start. Labels can change, so look for the same names in your account. **Can I put client data in project instructions?** Keep it to rules and firm documents. The glossary entry on [project instructions](https://learn.ynteractive.com/content/glossary/project-instructions) notes they are not a place for secrets or customer data, and the [ChatGPT data controls lesson](https://learn.ynteractive.com/content/chatgpt-data-controls-protect-client-data) covers the account settings to check. **Should I just use memory instead of writing instructions?** No. Memory is what ChatGPT chooses to save, and it can keep an old detail or miss a new one. Write the rules down where you can see and edit them. ## Free lessons in this post - [ChatGPT Projects: Keep Your Brand Voice, Files and Instructions in One Place](https://learn.ynteractive.com/content/chatgpt-projects-brand-voice): Set up a ChatGPT Project with your brand voice instructions, your price list and your style guide, so every new chat already knows your business. - [Turn a Great Prompt into Reusable Instructions](https://learn.ynteractive.com/content/turn-chatgpt-prompt-into-project-instructions): Take a prompt you paste every week, sort it into permanent rules and fill-in blanks, put the rules in a ChatGPT Project, and test it with a one-line request. - [Clear Client Data Out of ChatGPT Before It Becomes a Problem: Delete Chats, Clear Memory and Use Temporary Chat](https://learn.ynteractive.com/content/how-to-delete-chatgpt-history-clear-memory): Someone pasted client details into ChatGPT. Delete that one chat, archive what you keep, clear the one memory it left, set safe defaults and start using Temporary Chat. ## Watch it in a free lesson: ChatGPT Projects: Keep Your Brand Voice, Files and Instructions in One Place ### ChatGPT Projects: Your Brand Voice in One Place: video walkthrough Full walkthrough · 3:29 · with captions Set up a ChatGPT Project with your brand voice instructions, your price list and your style guide, so every new chat already knows your business. The whole lesson narrated step by step, recorded in Stepthrough’s simulated practice environment. [Watch it with chapters and the full transcript](https://learn.ynteractive.com/content/chatgpt-projects-brand-voice#watch) From the free lesson [ChatGPT Projects: Keep Your Brand Voice, Files and Instructions in One Place](https://learn.ynteractive.com/content/chatgpt-projects-brand-voice) ## Related terms Plain-English definitions, each with an example from a free lesson. [Custom instructions (ChatGPT)](https://learn.ynteractive.com/content/glossary/custom-instructions) · [Project instructions](https://learn.ynteractive.com/content/glossary/project-instructions) · [New chat](https://learn.ynteractive.com/content/glossary/new-chat) · [Archived chats](https://learn.ynteractive.com/content/glossary/archived-chats) · [Chat history](https://learn.ynteractive.com/content/glossary/chat-history) · [ChatGPT saved memories](https://learn.ynteractive.com/content/glossary/chatgpt-saved-memories) · [Persistent agent](https://learn.ynteractive.com/content/glossary/persistent-agent) · [ChatGPT memory](https://learn.ynteractive.com/content/glossary/chatgpt-memory) [The full AI glossary](https://learn.ynteractive.com/content/glossary) ## Related prompts - [ChatGPT memory sort prompt](https://learn.ynteractive.com/content/prompts/chatgpt-memory-sort-prompt): Paste your ChatGPT saved memories and sort each one: keep it, move it to custom instructions, or delete it because it holds… - [Firm project instructions draft prompt](https://learn.ynteractive.com/content/prompts/firm-project-instructions-draft-prompt): A prompt that turns your accounting firm's tone, fee rules and client policies into project instructions for Claude or ChatGPT… - [IT incident report prompt](https://learn.ynteractive.com/content/prompts/it-incident-report-prompt): A ChatGPT prompt for an IT incident report: turn your log into a timeline, what was affected, how you contained it and what is… [All free prompt templates](https://learn.ynteractive.com/content/prompts) ## Visual guides The same ideas as quick visual guides: swipe through the slides, or read each one as text. - [Mistake: Re-Explaining the Client in Every ChatGPT Chat](https://learn.ynteractive.com/content/visual/mistake-re-explaining-the-client-every-chat): A common ChatGPT mistake: explaining the same client again in every new chat. Give each client one Project with its own instructions and files. - [Myth: ChatGPT Memory Is Where Your Work Rules Belong](https://learn.ynteractive.com/content/visual/myth-chatgpt-memory-holds-your-rules): Myth vs fact: ChatGPT memory is not where work rules belong. Put standing rules in custom instructions and one client's rules in its Project. - [Mistake: Deleting a ChatGPT Chat but Keeping the Memory](https://learn.ynteractive.com/content/visual/mistake-deleting-the-chat-keeps-the-memory): A visual guide to a common ChatGPT cleanup mistake: deleting a chat with client details but leaving the saved memory behind. Delete the memory too. [All visual guides](https://learn.ynteractive.com/content/visual) ## Keep reading - [Does ChatGPT Remember Other Chats? Chat History vs Memories](https://learn.ynteractive.com/content/blog/chatgpt-reference-chat-history): ChatGPT can remember other chats through saved memories and the Reference chat history setting. What each one does, and how to control both at work. - [ChatGPT Memory vs Custom Instructions: What's the Difference](https://learn.ynteractive.com/content/blog/chatgpt-memory-vs-custom-instructions): Custom instructions are rules you write; memory is what ChatGPT picks up from your chats. See how each works, where to find them and which to use for work. - [ChatGPT Shared Project vs Shared Chat Link: Which to Use](https://learn.ynteractive.com/content/blog/chatgpt-shared-project-vs-shared-chat-link): A shared chat link shows one conversation; a shared project gives people your chats, files and instructions. See what each reveals before you send it. [All blog posts](https://learn.ynteractive.com/content/blog) · [All free tutorials](https://learn.ynteractive.com/content)