Home / Work

Make ChatGPT remember your project

To make ChatGPT remember a project, put the facts in a Project or in a short brief you paste every time. Saved memory, when your settings show it, keeps some details, not every word of every chat. End each session with a handoff note for the next one.

Next step: do Step 1. Copy the box, fill in the blanks, and paste it into the chat.

See saved guides

WorkChatGPT, Claude, or Gemini20 minutesUpdated

A prompt is the message you paste. A Director check is a short look before you use the result.

What do you need before you start?

Gather the items below before you open the chat. You already have most of them.

  • The chat you already use, signed in
  • One project, in the plain sense: a job, a class, a trip, a move
  • Ten minutes to write a brief you could hand to a coworker

You do not need a paid plan to use the brief and the handoff. Paid controls, when they exist, are extra. If a menu below is missing, skip that menu and keep the brief.

How do you see what this chat does and does not remember?

A single conversation knows what is in that conversation. When you close it and open a new one, do not expect the new one to know the old one unless a memory setting says so.

OpenAI's memory help says that when Memory is on, ChatGPT can use relevant details from sources such as past chats, saved memories, and custom instructions. It also says Memory does not keep every detail, and that features vary by plan, region, platform, and workspace. If a setting on that page is not on your screen, it may not be on your account yet.

Check it yourself:

  1. Open Settings.
  2. Select Personalization.
  3. Select Memory, if it is there.

You may see Memory, Reference saved memories, Reference chat history, or a memory summary. Those names come from OpenAI's help. Turn on only what you understand. You can ask what it remembers, correct a line, or delete remembered information. Turning Memory off does not delete your old chats.

Custom instructions are different. They are the standing note you wrote about how you want replies. Memory is what the product saves from use. Do not treat either one as a filing cabinet.

How do you use a Project when the sidebar shows one?

OpenAI's Projects help says Projects keep related chats, files, and instructions together, for signed-in users, subject to plan and workspace settings. The same page says you can choose default memory or project-only memory when that control is in the project settings.

Project-only memory, in that help, means chats in the project can use other chats in that project, and they do not use saved memories from outside it. Shared projects use project-only memory. If you do not see the switch, leave it. Check the help page rather than guessing your plan.

To make one:

  1. In the sidebar, choose Projects, then New project. Names vary. If Projects is not there, skip to Step 3.
  2. Name it for the work, not "stuff."
  3. Put the brief from Step 3 into the project instructions.
  4. Add a file only if you are allowed to store it there. OpenAI says the number of files depends on the plan. Check the plans page for the current limit.

A chat you start inside the project should follow those instructions. A chat you start outside it should not be trusted to know them.

Claude: the projects article says projects are available to free accounts, up to five, and that uploaded files are used across chats in that project. The how-to says to open Projects, or go to claude.ai/projects, then New Project. Add the brief as project instructions if that box is on the page. A later Claude help page talks about some instruction options on paid plans. Use the box you see.

Gemini: the Gems help says to open gemini.google.com, open the sidebar, choose Gems, then New Gem, enter a name and instructions, and save. Optional files sit under Knowledge. If Gems is not in the sidebar, you do not have that control on this account. Use the brief anyway.

How do you write a one-page project brief?

This is the part that works in every tool. Keep it short enough to paste. The chat should not have to guess the goal.

Project brief. Use this in every reply. Do not add facts.

Goal: [what done looks like]
Already decided: [three bullets you mean]
Still open: [questions]
Do not assume: [names, dates, numbers, or tools it must not invent]
Next action: [the one thing you will do after this chat]
Voice: [how you want the writing to sound]

Save that note on your computer. Paste it into project instructions if you have them. If you do not, paste it at the top of one long-running chat and keep working in that same chat until the job is done. Starting a fresh chat without the brief is how the project "forgets."

How do you end every session with a handoff?

Before you close the tab, make the next session cheap. This note is yours. The product's memory is a bonus.

Write a handoff I can paste into the next chat.
Include only what we actually decided in this conversation.
Use these headings:
- Decisions
- Facts I supplied (do not add new ones)
- Open questions
- Next action, one line
Stop. Do not start the next task.

Read it. Delete anything you did not say. Paste that handoff, plus the brief, at the start of the next chat. That is the memory you control.

For a job with many phases, do not ask for the whole thing in one message. Break a big job into phases, and ask for a plan first.

What should you keep private?

OpenAI says sensitive information can show up in Memory if you share it, and that you should review Memory and Sources. Do not save passwords, bank details, medical notes, ID numbers, or other people's private mail. Do not connect an app such as Gmail unless you want that mail available as context. A temporary chat, with personalization off, is the safer place for a one-time question. Do not put a client's confidential file in a personal Project. If this is work, follow the company rule.

What should you check before you use it?

Answer each line below. If one is still a no, do not use the result yet.

Director check: look before you use it
  • Can a brand-new chat finish the next step from the brief and the handoff alone? If not, the brief is too thin.
  • Did you check Settings instead of believing a screenshot from the internet? Plans change. Your screen is the source.
  • Is anything in Memory that you would not want repeated? Delete it, or turn Memory off.
  • Did you start the next chat inside the Project? A chat outside it does not inherit the instructions.

What is one extra tip?

Keep the brief next to the work, the way you would keep a checklist. When the job is a series of phases, use Break a big job into phases. When you are about to ask for the finished piece too soon, use Ask for a plan before the work.

What else do people ask?

Open a question. The short answer under it is the one to use.

Does ChatGPT remember everything from old chats?

No. OpenAI's memory help says Memory does not retain every detail from every conversation, and the controls you see depend on your plan, region, and workspace. Open Settings, then Personalization, then Memory. If a switch is missing, do not assume that feature is on your account.

What is a Project, and how is it different from a chat?

OpenAI describes a Project as a place that keeps related chats, files, and instructions together. A normal chat is one conversation. A Project can use its own chats and files as context for later replies in that Project. Sharing and file limits depend on the plan. Check OpenAI's plans page. This guide does not list prices.

What if I use Claude or Gemini instead?

Claude's help says Projects are available on free accounts, with a maximum of five projects, and that you can add files to the project knowledge. Gemini's help says you can save a Gem with a name and instructions when Gems shows in the sidebar at gemini.google.com. If you do not see those controls, use the same brief and the same handoff note in one long chat.

Can I stop it from saving private facts?

Yes. Do not type secrets into Memory. OpenAI says you can delete remembered information from the memory summary or saved memories, and that a temporary chat does not create memories. Claude and Gemini have their own settings. If you are unsure, leave the fact out.

By Joe Rangel · Last updated · More Work guides

Get one workflow like this every morning.

5-minute read. Real tasks. Zero hype.

More Work guides

See all