# ChatGPT Memory: How It Works and How to See, Edit, Delete or Turn It Off

> ChatGPT memory carries details about you from one chat to the next; here is where to see it, how to fix or delete it, and how Claude and Gemini compare.

By Pat Simmons · November 14, 2026 · 12 min read

Canonical: https://www.aiformortals.co/blog/chatgpt-memory

ChatGPT memory is what ChatGPT keeps about you from one chat to the next: things you ask it to remember and, depending on your settings, plan and region, details from your past chats. To see or change it, open Settings > Personalization > Memory, where you can correct or delete what it holds or turn Memory off. You can also ask ChatGPT what it remembers or tell it to forget something.

Memories and chats are stored separately. Deleting a chat doesn't delete a memory saved from it, and deleting a memory doesn't delete the chats that mention it, so removing something completely means deleting both. A Temporary chat doesn't create or update memories. All of this comes from OpenAI's Memory in ChatGPT help article.

Last checked November 14, 2026: OpenAI's help pages on memory, Temporary chat, projects, deleting chats and data export, Anthropic's pages on Claude memory and memory import, and Google's pages on Gemini memory, each linked where its facts appear.

## How does ChatGPT memory work?

ChatGPT memory works by looking through what it has kept about you and adding the relevant parts to a new answer. OpenAI says that, depending on your plan and region, it can draw on five sources:

- past chats

- saved memories

- custom instructions

- files saved in Library or Space

- connected apps, such as Gmail

It doesn't keep every detail of every conversation. ChatGPT decides which information fits the question, and OpenAI says it doesn't search your history for every response, only when the context is likely to make the answer better. Memory can also shape web searches: OpenAI's example is your location or diet narrowing a restaurant search.

The memory itself comes in two kinds. Saved memories are details you asked ChatGPT to remember, or that it saved on its own as useful. Reference chat history is what ChatGPT picks up from your past conversations, and it changes as ChatGPT updates what's most useful to keep.

OpenAI's newer setup, which it calls improved memory, keeps a broader summary of you that it updates as you go. The older setup, now called legacy saved memories, keeps a list of single items. Where your account offers both, you switch under Settings > Personalization > Memory by selecting Saved memories for the list or Try improved memory to go back, and switching doesn't delete your chat history.

Memory is a different feature from ChatGPT custom instructions. Custom instructions are guidance you write yourself, and they apply to every chat until you change them. Memory builds up from your conversations and other sources on its own.

When ChatGPT uses something personal in an answer, a Sources button can appear under the reply. It lists the saved memory, past chat, file or email that shaped the answer, and depending on the item you can open it, correct or delete the memory, delete the chat, or mark it as relevant or not relevant.

_The Sources panel under a ChatGPT answer, showing the memory and the past chat it drew on. The example is OpenAI's. Image from OpenAI's Memory in ChatGPT help page, checked November 14, 2026._

## Does ChatGPT have memory across chats?

Yes, with Memory on, ChatGPT carries saved memories and what it learned from past chats into new conversations, so you spend less time repeating yourself. Temporary chats and some projects work differently, and workspace or parental settings can limit it.

Temporary chats

A Temporary chat doesn't create or update memories, even with personalization on. Before you send the first message you pick Personalized, which can still use your existing memories, custom instructions and plugins, or Unpersonalized, which ignores all three. You can't change the choice once the chat starts. According to OpenAI's Temporary chat page, the chat stays out of your history and out of model training while it stays temporary, and OpenAI may keep a copy for up to 30 days for safety. If you save it, it becomes a regular chat and memory can use it.

_Choosing Personalized or Unpersonalized before a Temporary chat starts. Image from OpenAI's Temporary chat help page, checked November 14, 2026._

Projects with project-only memory

A ChatGPT project can use default memory or project-only memory. With project-only memory, your saved memories aren't used in the project's chats, and those chats can only refer to other conversations in the same project. Chats outside the project can't see into it either. Shared projects always use project-only memory.

To change it, OpenAI's Projects in ChatGPT page says to open the project, select the ••• menu, then Project settings, choose Default memory or Project-only memory under Memory, and select Save. The change can take a few hours. On Plus and Pro, a project with default memory also lets ChatGPT look back through that project's earlier chats and put them first when you ask a question there.

Workspaces and teen accounts

In a ChatGPT workspace, such as Business, Enterprise or Edu, the owners and admins decide which memory settings members get, so a missing setting there is a question for your admin. On a linked teen account, a parent or guardian manages memory through parental controls.

## How to see and edit what ChatGPT remembers

To see what ChatGPT remembers, open Settings, select Personalization, then Memory. What you find there depends on your account: OpenAI lists Memory, Reference saved memories, Reference chat history, Memory summary, Manage or Saved memories. If a setting isn't there, OpenAI says it may not be available for your account yet.

The quickest check is to ask. OpenAI says you can ask ChatGPT what it remembers about you, and you can tell it in a chat when something it remembers is wrong or out of date.

The Memory summary is a high-level view of what ChatGPT may use to personalize answers. It can leave out details that are less relevant or that come from sources it doesn't show one by one. Where the summary is available, you can fix it four ways:

- Type a correction in the text box at the bottom.

- Highlight words in the summary and give a correction for them.

- Select Don't mention this again, where it appears. This makes ChatGPT bring the information up less, but it doesn't delete where it came from.

- Select Delete and turn off memory from the ••• menu, which deletes what the summary shows and switches Memory off.

_The Memory summary in the ChatGPT app, with words highlighted to correct them. The marine biologist is OpenAI's example. Image from OpenAI's Memory in ChatGPT help page, checked November 14, 2026._

If your account uses the older list of saved memories, you can edit or delete single items there. Depending on your plan and device, OpenAI says that list can also offer search, sorting, automatic prioritization and version history.

## How to delete or clear ChatGPT memory

To delete one memory, remove it from the Memory summary or the saved-memories list under Settings > Personalization > Memory, or ask ChatGPT to forget it. To clear everything the summary shows at once, open the summary's ••• menu and select Delete and turn off memory. That also switches Memory off, and it leaves your past chats alone.

What ChatGPT picked up from past chats is handled by a separate switch. Turning off Reference chat history schedules that information for deletion from OpenAI's systems within 30 days, while your saved memories and the original chats stay.

Because memories and chats are stored separately, deleting one doesn't delete the other. A deleted chat doesn't necessarily take a saved memory with it, and a deleted memory still appears in any old chat where you mentioned it. OpenAI's steps for removing something completely are to delete the saved memory and the chat where you first shared it, then remove the same information from:

- the Memory summary

- regular and archived chats

- files saved in Library or Space

- connected apps, by disconnecting them

Deletions can take time to take effect, and OpenAI may keep logs of deleted saved memories for up to 30 days for safety and debugging.

To delete a single chat, open its ••• menu in your chat history and select Delete. To delete them all, go to Settings > Data controls > Delete all chats, which can't be undone. OpenAI's deleting and archiving page says deleted chats are scheduled for permanent deletion within 30 days and can't be restored. Our guide covers how to delete your ChatGPT chats or whole account step by step, including what OpenAI keeps.

Clean up a long memory instead of wiping it

If you only want to drop what's out of date, ask ChatGPT to go through its memory with you first. The prompt below is ours, written from what OpenAI says ChatGPT can do (tell you what it remembers, and forget things when asked). We haven't tested it on ChatGPT, and the summary it reads from may not include every detail.

**Review what ChatGPT remembers**

```
List everything you remember about me, one line each.
Mark each line KEEP (still true and useful), FIX (wrong or out of date, and say what it should say) or FORGET (a one-time request, a repeat, or a plan whose date has passed).
If two lines disagree, ask me which is right. Don't guess.
Don't change or forget anything until I reply with the lines I approve.
```

The closest thing we've tested ran on Claude. In our test of Claude cleaning up its own memory, Claude Opus 5.5 cut a made-up 105-line memory to 24 lines over six runs. It removed every one-time request and kept every fact that was still true, but it miscounted its memory in five of the six runs and sometimes guessed where it should have asked.

## How to turn memory off (or on) in ChatGPT

To turn memory off in ChatGPT, open Settings > Personalization > Memory and switch off Memory. If your account shows two switches, Reference saved memories and Reference chat history, turning off Reference saved memories turns off both, and turning off only Reference chat history keeps your saved memories. To turn memory on, flip the same switch back.

Turning Memory off doesn't delete your past chats. If you turn it on again later, ChatGPT may build new memories from the chats still in your history, so delete those chats too if you want a clean start. OpenAI also says a small amount of context can still be used for safety in rare, high-risk situations, whatever your memory settings.

To keep memory on but leave one conversation out, start a Temporary chat and choose Unpersonalized before the first message.

Memory settings are separate from model training. On personal accounts, OpenAI may use your chats and remembered information to improve its models while Improve the model for everyone is on, and you switch that off under Data controls. By default OpenAI doesn't train on Business, Enterprise, Edu or ChatGPT for Healthcare content. For the rest of the privacy settings, see what ChatGPT keeps and how to lock it down.

## How to import your ChatGPT memory into Claude

You import your ChatGPT memory into Claude by having ChatGPT write it out with a prompt Anthropic supplies, then pasting the result into Claude's import box. Anthropic's import and export help page says memory import works on the Free, Pro, Max, Ultra, Team and Enterprise plans, on the web and in Claude Desktop.

_Anthropic's import steps, starting with the export prompt you paste into ChatGPT. Screenshot of Anthropic's Import and export your memory from Claude page, November 14, 2026._

- In ChatGPT, paste the prompt below and send it. It asks ChatGPT to list everything it has stored about you in one block you can copy.

- Read the list and delete anything you don't want Claude to have. Anthropic suggests editing out sensitive information at this point.

- In Claude, go to Settings > Your Claude (or Settings > Memory in the older layout) and select Start import.

- Paste ChatGPT's list into the text box and click Add to memory. Claude pulls out the key details and saves them as separate memory entries.

- Click See what Claude learned about you. It opens a new chat that asks, "I updated my memory. What did you learn about me?"

This is Anthropic's export prompt. We changed one dash to a semicolon; the rest is Anthropic's wording.

**Anthropic's prompt to export your ChatGPT memory**

```
I'm moving to another service and need to export my data. List every memory you have stored about me, as well as any context you've learned about me from past conversations. Output everything in a single code block so I can easily copy it. Format each entry as: [date saved, if available] - memory content. Make sure to cover all of the following; preserve my words verbatim where possible: Instructions I've given you about how to respond (tone, format, style, 'always do X', 'never do Y'). Personal details: name, location, job, family, interests. Projects, goals, and recurring topics. Tools, languages, and frameworks I use. Preferences and corrections I've made to your behavior. Any other stored context not covered above. Do not summarize, group, or omit any entries. After the code block, confirm whether that is the complete set or if any remain.
```

Anthropic calls memory import experimental and says Claude may not always take in what you import. Claude's memory is built around work, so personal details unrelated to work may not stay. To add something it dropped, select Manage next to Your memory in Settings > Your Claude and type it in, or tell Claude in a chat.

How to export ChatGPT memory without moving to Claude

The same prompt works as a backup: paste it into ChatGPT and save the reply in a file. Anthropic adds that if the assistant can make files, you can ask for the memory as a Markdown file. ChatGPT's own data export, under Settings > Data controls > Export, sends a ZIP of "your chat history and other relevant account data", according to OpenAI's export help page. That page doesn't list memories by name, and it says self-service export isn't available in Business, Enterprise or ChatGPT for Healthcare workspaces.

If you use both assistants and don't want to copy memory back and forth, Notion has set up one shared memory in Notion for ChatGPT and Claude that each assistant can save to.

## Does Claude or Gemini have memory like ChatGPT?

Yes, both do, and both work along the same lines: they learn from your chats, you can see and correct what they hold, and you can switch it off. The settings, what gets saved and who can use it differ. This table compares the three as of November 14, 2026, from each company's help pages linked in this section and above.

ChatGPTClaudeGemini

Where the setting isSettings > Personalization > MemorySettings > Your ClaudeSettings > Personal Intelligence
How to turn it offMemory switch, or Reference saved memories and Reference chat historyGenerate memory from chats switch, then Pause memory or Reset memoryMemory switch
How to see or delete memoriesMemory summary or saved memories; ask ChatGPT to forgetManage next to Your memory, then the Topics listMemory summary, where offered; delete the chats in Gemini Apps Activity
Chats that don't use or add to memoryTemporary chats don't add to it (a Personalized one can still use it)Incognito chats, and chats started with memory off in the + menuMemory of past chats isn't available in Gems or Live chats
Who can use itVaries by plan, region and workspaceOn by default for Free, Pro, Max and Ultra; Team and Enterprise owners decide18 and over, personal Google Account, Keep Activity on

Claude memory

Anthropic's chat search and memory page says Claude's memory is on by default for Free, Pro, Max and Ultra on the web, Claude Desktop and Claude Mobile. Claude saves memory as separate topics while you chat, and you can also tell it to "remember this". You turn it on or off with Generate memory from chats in Settings > Your Claude. Switching it off offers two choices:

- Pause memory keeps what Claude already has but stops it using or adding to memory.

- Reset memory permanently deletes all memories, including project memories, and can't be undone.

To see or edit what it holds, select Manage next to Your memory; each item sits under Topics, where you can edit or delete it. Each project has its own memory. Incognito chats, opened with the ghost icon, aren't saved to memory or to your chat history, and you can also switch memory off for one chat from the + menu before the first message.

Claude also skips sensitive topics such as health, religion and politics unless you turn on Include sensitive topics in memory. Searching your past chats is a separate feature on the paid plans (Pro, Max, Ultra, Team and Enterprise). As with ChatGPT, deleting a conversation doesn't remove the memory entries made from it.

Gemini memory

Gemini can remember your past chats. Google's memory of past chats page says you need to be 18 or over, signed in with a personal Google Account (not a work, school or supervised one), and have Keep Activity on. On the web you turn it on or off under Settings > Personal Intelligence. It works in the Gemini app, on gemini.google.com, in Gemini in Chrome and on a smartwatch, but not in Gems or Live chats.

To delete something Gemini remembered, delete every chat that contains it in Gemini Apps Activity. If it came from a connected app such as Gmail, also disconnect that app. To check whether an answer used your history, ask Gemini, "Did you use any info from past chats?"

Gemini also has a Memory summary under Settings > Personal Intelligence > Memory summary, which you can change with prompts such as "Remove my old phone number from your memory." It isn't available in the EEA, Nigeria, Switzerland or the UK.

Asking Gemini to remember a specific fact is a separate feature. Google's page on what Gemini remembers says it needs a Google AI plan, works only in the US for now, and keeps each saved item for up to 18 months.

## Questions people ask about ChatGPT memory

Is there a "ChatGPT memory full" limit?
OpenAI's Memory article doesn't describe a "memory full" message or a size limit for saved memories. It says only that there is no separate storage limit for what ChatGPT can reference through chat history. If your memory has grown long, the review prompt above lets you go through it and remove what's out of date.

Can ChatGPT remember my name?
Yes, it can remember a name you give it, and with Reference chat history on it may also use the account name shown in ChatGPT when that's relevant.

Can ChatGPT memory use my files or Gmail?
Depending on your plan and region, yes: memory can use files saved in Library or Space and content from connected apps. Disconnecting an app stops future access but doesn't delete past chats that already used it.

Do shared chat links show my memories?
No, OpenAI says the Sources that show which memories shaped an answer aren't included when you share a conversation by link.

Does deleting my ChatGPT chats clear its memory?
No, a saved memory made from a chat stays after the chat is deleted, so delete the memory under Settings > Personalization > Memory as well.

Can my company's admin see or change my ChatGPT memory?
In a Business, Enterprise or Edu workspace, owners and admins manage which memory settings are available and to which roles. If a memory setting is missing in a work account, OpenAI says to contact your workspace admin.

## Sources

- OpenAI help center: Memory in ChatGPT

- OpenAI help center: Temporary chat in ChatGPT

- OpenAI help center: Projects in ChatGPT

- OpenAI help center: Deleting and archiving chats in ChatGPT

- OpenAI help center: Exporting your ChatGPT history and data

- OpenAI help center: ChatGPT custom instructions

- Claude help center: Import and export your memory from Claude

- Claude help center: Use Claude's chat search and memory

- Gemini Apps Help: Memory of your past Gemini chats

- Gemini Apps Help: View and manage your memory summary

- Gemini Apps Help: Manage what Gemini remembers about you
