How to Find ChatGPT Memory: View, Turn Off, Delete

7 min read

How to find ChatGPT memory in Settings, see what it remembers, turn memory on or off, and fully delete saved memories and chat history, on web and phone.

To find ChatGPT memory, open Settings > Personalization > Memory. There you’ll see a memory summary or a list of saved memories, plus switches to turn memory and chat history reference on or off. You can also just ask ChatGPT “What do you remember about me?” or tap Sources under a reply.

What is ChatGPT memory?

ChatGPT memory is the feature that lets ChatGPT reuse details from earlier chats to personalize new ones. Depending on plan and region, OpenAI says it can draw on past chats, saved memories, custom instructions, files in your Library or Space, and connected apps such as Gmail (Memory in ChatGPT, OpenAI’s help page, re-checked November 2, 2026).

It isn’t a full transcript. OpenAI states that memory “does not retain every detail from every conversation” and that ChatGPT picks what seems relevant to each answer. It’s also separate from your chat history in the sidebar, which is just the list of past conversations.

Memory is available on Free as well as paid plans. Free users got saved memories first and, from June 2025, references to recent conversations (ChatGPT release notes).

How to see what ChatGPT remembers

  1. Open Settings.
  2. Select Personalization.
  3. Select Memory.
  4. Read the Memory summary if your account has it, or select Manage (or Saved memories) to see the list of individual memories.
  5. In any chat, tap Sources under a personalized reply to see which saved memory, past chat, custom instruction or file shaped it.
  6. Ask ChatGPT directly: “What do you remember about me?”

The labels differ by account, because OpenAI is moving users between two memory systems (see below). If a setting is missing, OpenAI says it may not be available for your account yet.

The summary isn’t everything. OpenAI’s FAQ says it’s “a high-level view” that “may omit details,” so Sources and asking ChatGPT can surface things the summary doesn’t show.

Improved memory vs legacy saved memories

In June 2026 OpenAI started moving users to improved memory, starting with Plus and Pro in the US, with Free and Go to follow. It replaced a flat list with a running summary.

Improved memoryLegacy saved memories
What you seeA memory summary you can edit in placeA list of individual memory items
How it updatesUpdated automatically as your context changesSaved when you ask, or when ChatGPT judges a detail useful
How to correctType a correction, or highlight text and correct itEdit or delete a single item
Extra options“Don’t mention this again”Search, sorting, automatic prioritization, version history (plan dependent)

Where both are offered, you switch in Settings > Personalization > Memory: choose Saved memories for the legacy list, or Try improved memory to go back. Switching doesn’t delete your chat history.

How to turn ChatGPT memory on or off

Go to Settings > Personalization > Memory and use the toggles. Depending on your account, you’ll see one Memory switch, or two: Reference saved memories and Reference chat history.

How they interact, per OpenAI:

  • Turning off Reference saved memories also turns off Reference chat history.
  • Turning off Reference chat history keeps your saved memories, and schedules what ChatGPT learned from past chats for deletion within 30 days.
  • Turning off Memory doesn’t delete past chats. If you turn it back on, ChatGPT may build new memories from chats still in your history.

For a single conversation without memory, start a Temporary Chat: open a new chat, select Temporary, and choose Unpersonalized before your first message. Since August 27, 2026 you can instead pick Personalized, which lets the temporary chat read your memories, custom instructions and plugins. You can’t change the choice once the chat starts. Either way, temporary chats never create or update memories, and they stay out of your history unless you save them. Saving turns it into a regular chat, so with memory on, ChatGPT may then use it later (Temporary chat help, release notes).

Two limits OpenAI states plainly: turning off Memory does not stop “limited safety and security uses of context in rare, high-risk situations”, and memory is separate from model training. On a personal account, chats and remembered information can train models while Improve the model for everyone is on. Turn that off in Data controls.

How to delete or wipe ChatGPT memory

Deleting memory takes more than one click because the same fact can live in several places. OpenAI lists what to clear:

To removeDo this
One memoryDelete it in Memory > Manage, edit the summary, or tell ChatGPT to forget it
All memory at onceIn the memory summary, open the ••• menu and choose Delete and turn off memory
The original sourceDelete the chat where you shared it, including archived chats
FilesDelete the file from Library (or Space, which replaces Library on eligible Pro, Business and Enterprise accounts since September 29, 2026)
Connected dataDisconnect the app (for example Gmail)

Three points catch people out:

  • Deleting a chat doesn’t delete a memory made from it. They’re stored separately.
  • Deleting a memory doesn’t edit old chats. Mentions stay in past conversations until you delete those too.
  • “Don’t mention this again” isn’t deletion. It reduces future references but keeps the source.

OpenAI may keep logs of deleted saved memories for up to 30 days for safety and debugging. On a phone, the steps are the same Settings > Personalization > Memory path. OpenAI noted in June 2026 that “Delete and turn off memory” reached the web first, with mobile following.

ChatGPT memory full: how to free up space

The “memory full” message applies to the legacy saved memories list, which has a capacity limit. To free space, open Memory > Manage and delete entries you don’t need, starting with outdated or duplicate ones.

OpenAI has reduced the problem twice. In October 2025 it began managing saved memories automatically for Plus and Pro on the web, “moving less important ones to the background” to help “avoid ‘memory full’ state.” The June 2026 improved memory gave Plus and Pro “twice as much memory capacity.” OpenAI also says there’s no separate storage limit for what ChatGPT can reference through chat history.

What changed in ChatGPT memory, August to October 2026

From OpenAI’s release notes and help pages, checked November 2, 2026:

DateChangeWhat it means for memory
Aug 14Change an existing project’s memory settingSwitch an unshared project between default and project-only memory; shared projects stay project-only
Aug 27Personalized temporary chatsA temporary chat can read your memory, but still never writes to it
Sep 29Space replaces Library (eligible Pro, Business, Enterprise)Files ChatGPT can draw on now live in Space; delete them there
Oct 20Locked chats (PIN or Face ID)Locked chats can use your regular memory, but new messages in them don’t personalize regular chats
Oct 30Account switching on iOS and AndroidUp to 2 accounts signed in; each keeps its own chats and memory

Locked chats and memory. Locking a chat hides it from your history and search. It doesn’t erase what ChatGPT already remembered from that chat. If you lock a chat that sits in a project, it leaves the project and stops using that project’s memory. And resetting your account PIN permanently deletes all locked chats and any memory stored only inside them; OpenAI Support can’t recover it (Using locked chats).

Memory in ChatGPT projects

Projects can have project-only memory. With it on, chats in the project can use other chats in that project, but nothing from outside, and nothing from the project leaks into other chats. You can change an existing project’s setting from the project’s three-dot menu under Project settings > Memory. Shared projects always use project-only memory.

In ChatGPT for Healthcare and ChatGPT Enterprise with Regulated Workspace, improved memory is off by default and admins decide who gets it. OpenAI says it isn’t covered under the BAA, so don’t enter health data (PHI) while using it.

This is the cleanest way to keep, say, a work project from coloring personal chats. Claude also has per-project memory, but by default it mixes in your regular-chat memory unless you turn off Use account memory (Claude’s project memory); see how chatbot memory compares across apps.

ChatGPT memory vs memory you build yourself

ChatGPT’s memory is a feature of the ChatGPT app. If you build on OpenAI’s developer tools, you design memory yourself; memory in the OpenAI Agents SDK covers that side. Open-source memory servers such as Hindsight give your own agent a store you can inspect and delete from, with an OpenAI Agents SDK integration; each write calls an LLM, so it costs more than a plain database. For why every chatbot still forgets things, even with memory on, see why AI memory is so bad, and for Anthropic’s version, Claude’s long-term memory.