ChatGPT Memory: How It Works, How to See It and How to Turn It Off

ChatGPT memory lets ChatGPT remember details about you across conversations, such as your job, your preferences or an ongoing project, so you don’t have to repeat them. It works in two ways: saved memories, which you can view and delete one by one in ChatGPT’s personalization settings, and referencing past chats, where ChatGPT draws on your earlier conversations. You can switch either off, delete individual memories, clear them all, or use a temporary chat when you don’t want anything remembered.

Memory is useful, and it’s also the ChatGPT feature people are most often surprised by. This guide explains what it stores, how to check and control it, and what it doesn’t do, including the common worry that ChatGPT will repeat your details to other people.

How ChatGPT memory works

OpenAI describes memory as two related features. Availability and naming can vary by plan and region, and OpenAI has changed them over time, so check what your own settings show.

Feature	What it does	Can you see what’s stored?	How to control it
Saved memories	Specific facts ChatGPT keeps, either because you asked it to or because it judged them useful	Yes, as a list in memory settings	Delete items one by one, clear all, or turn saved memories off
Referencing chat history	ChatGPT draws on your past conversations to shape new answers	Not as a list; it’s based on the chats in your history	Turn it off, or delete the chats you don’t want it to draw on
Custom instructions	Standing preferences you write yourself, such as tone or what to call you	Yes, you wrote them	Edit or clear them in personalization settings

When ChatGPT saves something new to memory, it usually shows a small note in the conversation that memory was updated. That note is your cue to check what was saved if it matters to you.

What ChatGPT memory might remember about you

Typical saved memories look like short notes:

• “Works as a real estate agent in Denver.”
• “Prefers short answers with bullet points.”
• “Is planning a trip to Japan in the spring.”
• “Has two children and a dog named Max.”
• “Is writing a novel set in the 1920s.”

Most are harmless and make answers more useful. The concern is when memory quietly collects things you wouldn’t have written down deliberately: a health condition, a conflict with a colleague, a client’s name, a legal worry. OpenAI has said it tries to steer ChatGPT away from proactively remembering sensitive information such as health details unless you ask it to, but you shouldn’t rely on that. Check the list.

How to see what ChatGPT remembers

1. Open ChatGPT’s settings from your profile menu.
2. Go to the personalization section.
3. Open the option to manage memories. You’ll see the list of saved memories.
4. Read through it. Delete anything that’s wrong, outdated, or more personal than you’re comfortable with.

You can also ask ChatGPT directly, “What do you remember about me?” The answer is a helpful summary, but the settings list is the authoritative view of saved memories.

How to delete ChatGPT memories

Delete one memory

In the memory list, use the delete option next to the item. You can also tell ChatGPT in a chat to forget something specific (“Forget that I work at Acme”), then check the list to make sure it’s gone.

Clear all memories

The memory settings include an option to clear all saved memories at once. Use it if you’ve shared a lot over time and want a fresh start.

Remove what chat history referencing draws on

Referencing past chats works from the conversations in your history. Deleting a chat stops it being used this way, and turning the feature off stops it altogether. Our guide on how to delete ChatGPT history covers deleting chats individually or all at once.

How to turn ChatGPT memory off

In personalization settings, you’ll find separate switches for saved memories and for referencing chat history. You can turn off one, both, or neither.

• Turning off saved memories stops ChatGPT saving new memories and using existing ones. Depending on the current design, the existing list may stay stored until you delete it, so clear it too if you want it gone.
• Turning off chat history referencing stops ChatGPT drawing on past conversations. Your chats stay in your history unless you delete them.

If you only occasionally want a conversation kept out of memory, you don’t need to turn anything off. Use a temporary chat instead.

Temporary chats and memory

A temporary chat is ChatGPT’s “incognito” option. It doesn’t use your existing memories, doesn’t create new ones, doesn’t appear in your history and isn’t used to train models. OpenAI says it may keep a copy for a limited period for safety reasons; its help pages give the current details.

Temporary chats suit one-off sensitive conversations: a medical question, a draft about a family issue, anything involving someone else’s personal details. You lose personalization for that conversation, which is usually the point.

What ChatGPT memory doesn’t do

• It doesn’t share your details with other users. Your memories belong to your account. When someone else asks ChatGPT about you, it doesn’t consult your saved memories.
• It isn’t the same as training. Memory personalizes your own experience. Whether your content can be used to improve OpenAI’s models is a separate setting in data controls.
• It isn’t a reliable record. Memories can be wrong or out of date, and ChatGPT can misapply them. If it keeps assuming something that’s no longer true, delete the memory.
• It doesn’t make stored data disappear from OpenAI. Memories are data OpenAI holds under its privacy policy. Deleting them follows the same deletion process as other content.

The first point is worth repeating because it causes a lot of worry. If you’re concerned about what ChatGPT says about you to other people, memory isn’t the cause. That comes from what’s published about you online. Our guide on managing your reputation in AI search covers that, and our guide to AI hallucinations explains why AI tools sometimes invent details about real people.

A worked example

This is an illustrative scenario, not a real client. A school counselor uses ChatGPT to help plan workshops. After a few months she opens the memory list out of curiosity and finds, alongside useful notes about her job, a memory that says a named student “is struggling after a parent’s divorce”. She had mentioned it once while asking for advice on a lesson.

She deletes that memory straight away, then deletes the conversation it came from. She keeps memory on, because the notes about her workshops save her time, but she starts using temporary chats whenever a question involves a specific student, and she removes names before asking anything. She also checks her school’s guidance on AI tools, which says student information shouldn’t go into personal accounts at all. Memory itself wasn’t the problem; the detail she typed was.

Memory in projects and work accounts

ChatGPT projects can keep context within a project, so chats and files grouped together can inform each other. If you use projects for sensitive work, review what’s inside them as well as your main memory list.

On work or school workspace plans, administrators may control whether memory is available and how data is retained. If you’re not sure how your organization handles it, ask your IT team, and don’t assume personal-account settings apply.

Should you keep ChatGPT memory on?

It depends on how you use ChatGPT. A simple way to decide:

• Keep it on if you use ChatGPT often for the same kinds of tasks, like writing in a consistent style, and you’re willing to review the memory list now and then.
• Turn it off if you share a device, use ChatGPT mainly for sensitive topics, or don’t want to think about what’s being saved.
• Use it with temporary chats if you want personalization for routine work but privacy for the occasional sensitive question.

Whatever you choose, the broader habits in our ChatGPT privacy guide matter more than the memory switch: don’t paste other people’s personal data, secure your account, and turn off training if that concerns you.

If your worry is really about your wider online footprint, not just one tool, our guide to cleaning up your digital footprint is a good next step, and our personal reputation management team can help with what others can find about you.

Common mistakes

• Never checking the list. Memory updates quietly. A quick review every so often catches surprises.
• Deleting the chat and not the memory. They’re stored separately.
• Turning memory off and assuming everything is gone. Clear existing memories too if you want them removed.
• Confusing memory with training. They’re controlled by different settings.
• Telling ChatGPT to “forget” and not checking. Confirm in the memory list that the item was actually removed.

Frequently asked questions

Does ChatGPT remember everything I say?

No. With saved memories on, it keeps selected details, which you can see in a list. With chat history referencing on, it can draw on past conversations. You can turn off either feature and delete what’s stored.

Can other people see my ChatGPT memories?

Other ChatGPT users can’t. Anyone with access to your account can, and on work or school plans, administrators may have some access under their organization’s settings.

How do I make ChatGPT forget something?

Delete the specific memory in personalization settings, or ask ChatGPT to forget it and then confirm it’s gone from the list. If the detail is in a past chat and history referencing is on, delete that chat as well.

Does a temporary chat use my memories?

No. Temporary chats don’t use existing memories or create new ones, and they don’t appear in your history.