How to make ChatGPT remember your project between chats
Put the work in a ChatGPT project, choose its memory setting on purpose, and keep your decisions in a written log that every new chat reads, instead of hoping memory picks them up. OpenAI is explicit that memory “does not retain every detail from every conversation”: ChatGPT decides what is relevant. A decision log saved in the project is the part you control.
What ChatGPT keeps on its own
| What | How it works | Good for |
|---|---|---|
| Saved memories | Details you ask ChatGPT to remember, or that it saves on its own as useful context. They are stored separately from chat history. | Stable facts: your role, clients, writing preferences. |
| Chat history | When Reference chat history is on, ChatGPT uses relevant information from past chats. OpenAI notes this “can change as ChatGPT updates what is most useful to remember.” | Picking up loose threads. Not for decisions you need exactly. |
| Project instructions | Apply only inside the project and “override your global custom instructions.” | Rules for how ChatGPT should work on this project. |
| Project sources | Files you upload and responses you save to the project. | The current brief and the decision log. |
To see your memory controls, open Settings → Personalization → Memory. The labels differ between accounts, so you may see Memory, Reference saved memories, Reference chat history or Saved memories.
Pick the right project memory setting
| Setting | What chats in the project can use | Choose it when |
|---|---|---|
| Default memory (Free, Go, Plus, Pro, Business) | Your saved memories, other chats in the project, and conversations outside it, unless another project uses project-only memory | It is your own project and you want ChatGPT to use what it knows about you. |
| Default memory (Enterprise, Edu) | Your saved memories and chats in the same project. “Chats remain contained within the project.” | The same reasons, on a work plan. |
| Project-only memory | Chats and files in this project only. “Your previously saved memories are not referenced in project chats.” | Client work, confidential work, or when chats from elsewhere keep bringing in stale context. |
To change it, open the project, choose •••, then Project settings. Under Memory, choose Default memory or Project-only memory, and save. OpenAI says the change “may take a few hours”. Earlier, the setting was fixed when you created the project. OpenAI’s current article says you can change it later. Shared projects always use project-only memory and cannot be switched.
Keep a decision log that the next chat reads
Memory summarizes. A decision needs the date, the reason and who agreed, and a new chat cannot tell a decision you replaced from the current one. A short log covers all of that.
DECISION LOG: [project]
Updated: [date]
Current decisions
- [date] [decision]. Reason: [why]. Agreed by: [who]. Revisit if: [condition]
Superseded
- [date] [old decision]. Replaced on [date] by: [new decision]
Open questions
- [question] (waiting on [person or fact])
Store it in the project
- Ask ChatGPT to update the log at the end of a working session, then check it.
- Save that response to the project. OpenAI calls the option “Save to project” or “Add to project sources”, depending on your account. You can also upload the log as a file.
- Keep one current version. Delete or replace the old one so ChatGPT does not read two logs that disagree.
Each project holds a limited number of files: 5 on Free, 25 on Go and Plus, and 40 on Edu, Pro, Business and Enterprise.
Tell ChatGPT to use it
Add this to the project instructions:
At the start of each chat, read the latest DECISION LOG in this project's sources.
Treat "Current decisions" as settled unless I reopen one.
When we agree on something new, say "Decision to log:" followed by one line I can add.
If something I ask contradicts a logged decision, point it out before answering.
Where ChatGPT memory stops
- Other assistants cannot read it. Claude, Grok and your coding tools have no access to ChatGPT’s memory or projects. See how to share context between ChatGPT and Claude.
- It does not search everything every time. OpenAI’s answer to whether ChatGPT searches your history for every response is “No.”
- You cannot list what a project remembers. OpenAI says project memory “does not show a list of memories like personal memory.” Keep the decisions you rely on in a log you can read.
- Temporary chats leave nothing behind. They “do not create or update memories, even with personalization on.”
Connect a project record ChatGPT can read
If the same decisions need to reach Claude or a coding agent, keep them in a project tool that ChatGPT connects to. ChatGPT adds custom connections as apps or plugins, depending on your account, and what you can do with one depends on your plan.
| ChatGPT plan | Add a custom connection | Save through it |
|---|---|---|
| Business | Only admins and owners | Yes |
| Enterprise and Edu | Admins, owners and members they authorize | Yes |
| Pro | Yes | No, read and fetch only |
| Free, Go and Plus | Not listed in OpenAI’s current article | Not listed |
Custom connections work in ChatGPT on the web, not in the mobile apps. With Flori, ChatGPT reads the project brief and decision history. On a plan that allows updates, it saves new decisions to the project, and you check them in Flori. The ChatGPT setup guide has the steps.
Test it in a fresh chat
- Start a new chat inside the project.
- Ask: “What have we decided about [topic], and why? Quote the log entry you used.”
- Compare the answer with the log.
If it answers from general memory, quotes a superseded decision, or cannot find the log, check that the project instructions mention the log and that only one version is saved in the project.
Frequently asked questions
Do ChatGPT projects have their own memory?
Yes. A project can use default memory or project-only memory. With project-only memory, chats can use other chats in the same project but nothing outside it, and your saved memories are not used.
Can I see what a ChatGPT project remembers?
Not as a list. OpenAI says project memory does not show one. Keep the decisions you need in a log you can read.
Why did ChatGPT forget something from another chat?
Either the project uses project-only memory, or ChatGPT did not judge the earlier chat relevant. Memory does not keep every detail, so anything you need exactly belongs in the decision log.
Does deleting a chat delete what ChatGPT remembered from it?
Not necessarily. OpenAI says deleting a chat alone does not necessarily delete a separate saved memory created from that chat. Delete the saved memory as well.
Sources
Checked October 19, 2026. Vendors change these pages often, so check the source before you rely on a plan detail.