How to turn off ChatGPT memory, and make it forget what it already knows

The Morse teamPublished November 4, 20268 min read

To turn off ChatGPT memory, open Settings, select Personalization, then Memory, and switch Memory off, which changes how ChatGPT answers but deletes nothing it already knows. OpenAI says that if you turn Memory back on later, "ChatGPT may create new remembered information from chats that remain in your history." To make it forget, choose Delete and turn off memory, turn off Reference chat history, and delete the chats the memories came from.

We read OpenAI's help pages on memory, temporary chat, projects, locked chats, data controls, chat retention, and Space, its release notes, and its pricing page on November 4, 2026. Several of those pages had been updated in the previous four days, and OpenAI says memory controls "can vary by plan, region, platform, and workspace settings," so a switch named here may be missing from your account. OpenAI earns credit for its memory page, which says in writing what each switch leaves behind, and each section below links to the OpenAI page it quotes.

How do I turn off memory in ChatGPT?

From one menu. OpenAI's memory page gives a single path, with no separate steps by device:

1. Open Settings.
2. Select Personalization.
3. Select Memory.
4. Turn off Memory. If your account shows separate switches instead, turn off Reference saved memories, which OpenAI says also turns off Reference chat history.

On the web, open your account menu to reach Settings. On iPhone and Android, open the sidebar and tap your profile icon to open Settings, as OpenAI's data controls page describes, then follow the same steps. OpenAI's memory page doesn't list a separate path for the Mac or Windows apps.

If you're signed in to two accounts in the mobile app, turn it off in each. Since October 30, 2026, each account keeps "its own chats, memory, settings, and subscriptions" (OpenAI release notes).

What's the difference between saved memories and reference chat history?

Saved memories are notes. Reference chat history is ChatGPT reading your past chats. OpenAI's memory page describes saved memories as "details that you explicitly ask ChatGPT to remember or that ChatGPT saves as useful context," stored "separately from chat history." Reference chat history lets ChatGPT "use relevant information from past conversations," and what it draws from them "can change as ChatGPT updates what is most useful to remember." There is "no separate storage limit" on it.

The line has blurred. OpenAI's post on its memory system says that since April 2025 a background process it calls dreaming has curated memories "automatically" from your chat history, and a rebuilt version that rolled out from June 4, 2026 keeps memories "updated automatically," in the release notes' words. The memory page now calls this Improved memory, and you can switch back to the older list of individual items by selecting Saved memories on the same Memory screen. Switching doesn't delete your chat history.

The two switches are linked in one direction only. Turning off Reference saved memories turns off Reference chat history too, but turning off Reference chat history "does not delete your saved memories."

How do I delete ChatGPT memory?

Open the Memory summary, select the more options menu (•••), and choose Delete and turn off memory. OpenAI says this "deletes the remembered information shown in the summary and turns Memory off," but "does not delete past chats," and it reached every user on the web on June 12, 2026, with mobile "soon," according to the release notes.

To clear the rest, turn off Reference chat history and delete single memories from the summary, the saved-memory list, or Sources under a reply. Step 4 of how to delete ChatGPT history permanently walks through both, and the rest of that guide clears the chats, archives, and files a memory can come from.

Does ChatGPT remember deleted chats?

It can. OpenAI's memory page says "deleting a chat alone does not necessarily delete a separate saved memory created from that chat," so delete the memory as well as the chat, as how to delete ChatGPT history permanently sets out.

Every ChatGPT memory control, what it deletes, and how long it's kept

Built from OpenAI's help pages and release notes as read on November 4, 2026. "Not stated" means we couldn't find a period on OpenAI's pages.

How do I use ChatGPT without memory for one chat?

Start a temporary chat and choose Unpersonalized before your first message. OpenAI's temporary chat page says an Unpersonalized chat "does not use memory, custom instructions, or plugins," while a Personalized one can read your existing memories, and you can't switch once the chat starts. Neither kind creates memories, unless you save the chat, which turns it into a regular chat that memory can learn from. What a temporary chat still leaves with OpenAI is in is ChatGPT private.

How do I keep ChatGPT memory inside one project?

Set the project to project-only memory. Open the project, select the more options menu (•••), then Project settings, choose Project-only memory under Memory, and select Save (OpenAI). Chats in the project can then draw on each other but not on your saved memories, and chats outside can't draw on the project. OpenAI says switching removes the project's information "from memory used outside the project," and changes "may take a few hours."

Since August 14, 2026, you can switch an existing project too, on every plan, according to the release notes. Shared projects are always project-only and can't be switched back.

Where can ChatGPT still remember you after you turn memory off?

In four places, according to OpenAI's own pages:

• Safety context. Turning off Memory "does not disable limited safety and security uses of context in rare, high-risk situations," as is ChatGPT private explains.
• Other people's ChatGPT, in Space. Space, OpenAI's shared pages for Pro, Business, and Enterprise, warns that "if a collaborator has Memory enabled, their ChatGPT may remember information from the shared page even when your Memory is off" (OpenAI).
• Locked chats. Locking a chat "does not remove information that ChatGPT remembered before it was locked." Memories made inside locked chats stay there and don't join regular memory when you unlock. Whether locking encrypts anything is in are ChatGPT locked chats encrypted.
• Your chats. They're the raw material. As long as they stay in your history, turning Memory back on can rebuild what you switched off.

Is ChatGPT memory used for training?

Yes, on personal accounts with Improve the model for everyone on, where OpenAI's memory page says it "may use chats and remembered information to improve models." Turning off memory leaves that switch as it was, and how to turn training off, with what else feeds it, is in does ChatGPT train on your data.

Does memory work differently on Free, Plus, and Pro?

The switches are the same, but how much ChatGPT remembers isn't. On November 4, 2026, OpenAI's pricing page listed "Limited memory and context" for Free, "Longer memory" for Go at $8 a month, "Expanded memory and context" for Plus at $20, and "Maximum memory and context" for Pro, from $100.

The June 2026 memory system reached Plus and Pro in the US first, with "twice as much memory capacity," and was to follow for Free and Go over the next weeks. Plus and Pro memory can also draw on files and a connected Gmail account, according to the release notes of May 5, 2026. In ChatGPT for Healthcare and Enterprise with Regulated Workspace, improved memory is off by default, and in any managed workspace the admins manage which memory settings you get.

How Morse keeps your chats from turning into a profile

Every control above works on copies OpenAI holds: your chats, the memories drawn from them, and the logs of what you deleted, all on OpenAI's servers under your account. Turning Memory off pauses the copying, and your chats stay there for ChatGPT to learn from again. Morse stores your chat history in your browser instead, encrypted, and our servers don't keep messages or log what you write.

• No chat archive on our servers. Your history is locked when you sign out, and because we don't store chats, we usually have very little to give authorities.
• No training switch to find. Morse doesn't train AI on your content, sell your data, or show ads.
• A temporary chat isn't saved. Morse's privacy policy says temporary chats "are kept in memory only and never stored," and attached files are sent with the message and never stored by Morse.
• Deleting means deleted. Clearing your browser or deleting your account deletes your chats. Morse can't recover them either, so keep a copy of anything you need.

Then pick who sees each message on its way to the model:

• Anonymous, to ask GPT something without it landing in a profile under your name. Our proxy removes your account, email, and IP address, and the provider gets the message without knowing it's you. It may keep the message for a limited time to check for abuse, under its own policy, so leave out details that identify you.
• Private, for chats you'd rather no host kept at all. Open-weight models like DeepSeek, Kimi, and GLM, plus Grok, run on hosts that agreed to keep nothing: Morse has zero data retention agreements with its inference, search, and other providers.
• E2EE, for what no one else should read. Your browser checks the enclave's proof that it's genuine, then seals the message so only that enclave can read it, and our servers relay ciphertext they can't read. It rests on the enclave hardware and attestation working as designed.

You can send a first message from the Morse home page without an account, or start free. End-to-end encrypted chats come with Pro, at $19 a month.