How ChatGPT uses your other chats
ChatGPT doesn’t reread every old conversation each time you type. OpenAI’s Memory help page says it “looks for relevant context when it is likely to improve a response,” and that Memory “does not retain every detail from every conversation.”
There are two separate things at work, and people often mix them up.
Saved memories are specific notes. You can ask ChatGPT to remember something (“I’m vegetarian”), or it may save a detail it thinks will help. They’re stored apart from your chats. Deleting the chat where you said something doesn’t delete a saved memory made from it.
Reference chat history lets ChatGPT draw on past conversations directly, without you asking it to remember anything. OpenAI says there’s no separate storage limit for what it can reference this way. What it pulls from can change over time as ChatGPT decides what’s useful.
Depending on your plan and region, Memory can also draw on custom instructions, files in your Library and connected apps such as Gmail (OpenAI, as of October 2026).
When ChatGPT does use something personal, a Sources note may appear under the reply. It can point to the past chat, saved memory or file that shaped the answer. OpenAI says this may not show every factor, so treat it as a hint rather than a full audit.
You can also just ask it: “What do you remember about me?” The memory summary in settings gives a high-level view too, but OpenAI says the summary doesn’t include everything.
How to turn it off
You can switch off chat history alone, or all of Memory. Here’s what each setting does, per OpenAI’s help pages as of October 2026:
| Setting | What it stops | What it doesn’t touch |
|---|---|---|
| Reference chat history off | ChatGPT using past chats; remembered info is scheduled for deletion within 30 days | Your saved memories and the chats themselves |
| Reference saved memories off | Saved memories, and it also switches off Reference chat history | Your chats stay in your history |
| Memory off | All memory personalization | Past chats aren’t deleted; turning Memory back on can create new memories from them |
| Temporary chat | That conversation creating or updating memories | Pick Unpersonalized first, or it can still use existing memories |
To get there: open Settings, select Personalization, then Memory. OpenAI notes that the exact controls vary by account. You might see Memory, Reference saved memories, Reference chat history, Memory summary, Manage or Saved memories. If a setting isn’t there, it may not have reached your account yet.
There’s also Delete and turn off memory in the memory summary’s ••• menu. It wipes what’s shown in the summary and turns Memory off, but your past chats stay.
One honest caveat. OpenAI says turning off Memory doesn’t disable “limited safety and security uses of context in rare, high-risk situations.” That’s separate from personalization, and you can’t switch it off.
To remove a specific detail for good, OpenAI says you need to delete it everywhere it lives: the saved memory, the chat where you first mentioned it (archived chats count), any Library file and any connected app. Logs of deleted saved memories may be kept for up to 30 days. Our guide on whether deleting ChatGPT history actually deletes it covers the full deletion picture.
Temporary chats and projects keep things apart
You don’t have to turn Memory off to keep one conversation separate. Two tools do it on a smaller scale.
Temporary chat. Open a new chat, select Temporary, and choose Personalized or Unpersonalized before your first message. Either way, the chat won’t create or update memories, won’t appear in your history and isn’t used for training while it stays temporary. Unpersonalized also ignores your existing memories and custom instructions. You can’t change the choice once you’ve started. OpenAI may keep a copy for up to 30 days for safety (OpenAI, as of October 2026).
If you save a temporary chat, it becomes a regular chat and follows your normal Memory settings from then on.
Projects with project-only memory. A project groups chats, files and instructions. Set it to Project-only memory (project ••• menu > Project settings > Memory) and chats inside can see each other but nothing outside. Chats outside can’t see in, and your saved memories aren’t used there. OpenAI says the change can take a few hours to apply.
With Default memory, it’s looser. On personal plans and Business, project chats can reference your regular chats and be referenced by them. Shared projects always use project-only memory, so collaborators never get your personal memories.
What other people can see
ChatGPT’s memory is one question. Whether a person can read your chats is another, and the answer depends on how you share and what kind of account you use.
Shared links. A link from a personal account shows the conversation as it was when you made it. Anyone with the link can open it, and you can’t limit it to named people or set an expiry. OpenAI says links are anonymous by default, though some older sharing versions may show the creator’s name, and memory Sources aren’t included. Manage or delete links in Settings > Data controls > Shared links > Manage. Deleting a link doesn’t remove copies someone already saved.
Shared projects. Members can see the project’s chats, files and member list, including names and email addresses. Your chats outside the project stay private.
Work and school accounts. This is where the answer changes. ChatGPT Enterprise and Edu workspaces can use OpenAI’s Compliance Platform, which feeds conversation logs into eDiscovery and data-loss tools. OpenAI says locked chats are included, and an account PIN doesn’t block authorized access through it. If your employer pays for your account, assume the organization can review what you type there. OpenAI says it doesn’t train on Business, Enterprise or Edu content by default, but that’s a separate question from admin access.
Model training. On personal accounts, OpenAI may use chats and memories to improve its models while Improve the model for everyone is on. You can turn it off in Data controls. That’s about OpenAI’s use, not about other users seeing your chats. On a personal account, other people only see a conversation if you share a link or a project with them.
If privacy questions about your devices are what brought you here, our guide to checking a laptop for a keylogger covers a different risk: someone capturing what you type before it reaches any app. And if ChatGPT stopped replying mid-conversation, see what happens when you hit the ChatGPT free limit.
What to do next
- Ask ChatGPT “What do you remember about me?” and open Settings > Personalization > Memory to see the summary and saved memories.
- Turn off Reference chat history if you don’t want past chats used. Keep saved memories if you like them.
- Use a Temporary chat set to Unpersonalized for anything you want kept out of memory entirely.
- Check Data controls > Shared links and delete any links you no longer need. Keep personal topics off your work account.
Sources
- Memory in ChatGPT, OpenAI Help Center
- Dreaming: Better memory for a more helpful ChatGPT, OpenAI, June 4, 2026
- Temporary chat in ChatGPT, OpenAI Help Center
- Projects in ChatGPT, OpenAI Help Center
- Sharing conversations and scheduled tasks in ChatGPT, OpenAI Help Center
- OpenAI Compliance Platform for Enterprise and Edu customers, OpenAI Help Center
More from Gadget Daily News
- What happens to your passkeys if you lose your phone?Usually nothing is lost. Passkeys saved to iCloud Keychain, Google Password Manager, Microsoft Password Manager or a password manager like 1Password are backed up and come back when…
- Does dialing *#21# tell you if your phone is hacked?No. Dialing *#21# only tells you whether your incoming calls are being forwarded to another number, and if so, which one. It can't see spyware, a hacked account…
- Someone asked for my Google Voice verification code: what now?If you haven't sent the code, don't. If you already have, it's fixable in a few minutes. The scammer most likely used it to set up a Google…
- Got a data breach notice letter? What to do nextDon't panic, and don't call any number in the letter until you've checked it's real. A breach notice means your information was exposed, not that anyone has used…