How to Turn Off ChatGPT Memory So It Stops Remembering
7 min read
To make ChatGPT stop remembering things, open Settings, go to Personalization, and switch memory off. That stops new memories, but it does not erase the old ones. To wipe what is already stored, delete the saved memories one by one or all at once, and for a single conversation that should leave no trace in your memory, start a Temporary Chat. The rest of this article covers the order to do those in, and the places where it quietly goes wrong.
This is the "how do I switch it off" guide. If you first want to understand what ChatGPT saves and why it sometimes keeps the wrong thing, read what ChatGPT's memory actually remembers and come back here when you have decided you want less of it.
What are you actually switching off?
Memory is not one thing, and that is the main reason people turn it off and are still surprised. OpenAI's help pages describe two ideas. Saved memories are details ChatGPT keeps as separate entries, either because you asked it to remember them or because it decided they were useful. Chat history reference is ChatGPT drawing on your earlier conversations to answer a new one.
The wording differs between versions of the help pages, and that matters. An older version describes two separate controls, "Reference saved memories" and "Reference chat history", and says that turning off the first also turns off the second. A newer version describes a single "Enable memory" switch under Settings, Personalization, Memory. OpenAI also says an improved memory experience is rolling out in stages across plans and countries. So the honest answer is that your Settings screen may show one switch or two. Look at what you actually see, and treat the names below as a guide rather than a script.
How do you turn off ChatGPT memory?
- 1
Open Settings and find Personalization
Open Settings from your profile menu, on the web or in the mobile app. The Personalization section holds the memory controls. The exact path can differ a little between the website and the apps, and OpenAI has said some memory management options were first limited to the web, so use the web if you do not see them in the app.
- 2
Switch memory off
Turn off "Enable memory", or, if your screen has two controls, turn off "Reference saved memories". If you also want ChatGPT to stop drawing on old conversations, turn off "Reference chat history" as well.
- 3
Check that it stuck
Start a new chat and ask what it remembers about you. A good answer is that it has no stored details. If it still names your job or your dog, you are probably looking at an old entry that was never deleted.
Two warnings from OpenAI's own wording are worth keeping in mind. Turning memory off does not delete your past chats, and if you turn memory back on later, ChatGPT may build new remembered information from the chats that are still in your history. OpenAI also says that turning memory off does not disable limited safety and security uses of context in rare, high-risk situations. For ordinary use that is not something you will notice, but it explains why the company does not promise that nothing is ever looked at.
How do you delete what ChatGPT already remembers?
Switching memory off freezes the list; it does not empty it. To remove entries, open Settings, Personalization, then the option to manage memories.
- To delete one memory, click the three dots next to it and choose Delete.
- To delete everything, use the three dots next to the search bar and choose Delete all.
- The memory summary, where your version has one, offers "Delete and turn off memory" in its own menu. That deletes the remembered information shown in the summary and turns the feature off in one move. It does not delete your past chats.
- You can also say "forget that" in a chat to remove one specific memory without opening Settings. Check the manage-memories list afterwards, because a spoken request is easy to get half right.
If you turned off chat history reference, OpenAI says information remembered from past chats is scheduled for deletion from its systems within 30 days. That is a schedule, not an instant wipe.
Why does ChatGPT still remember after I deleted the chat?
Because the chat and the memory are different objects. OpenAI states that deleting a chat does not remove a saved memory that came from it. To remove something completely, delete the saved memory in the manage-memories list and the chat where you first shared it. Doing only one leaves the other, and the leftover can bring the detail back.
This is the most common cause of "I deleted it and it knew anyway". The cleanup order that works is: delete the memory entry first, then the conversation, then check with a fresh chat. If you did it the other way round and the entry is still there, it is not too late; the entry is still in the list and can be removed now.
When should you use Temporary Chat instead?
Turning memory off is a permanent setting. A Temporary Chat is a one-conversation exception, which is better when you only occasionally want privacy, such as asking about a health worry or a salary negotiation without having it shape later answers.
According to OpenAI's help pages, a Temporary Chat stays out of your history unless you choose to save it, and it does not create or update memories. Depending on the option you pick when you start it, the chat can read your existing memories or not, and you cannot change that choice after the conversation has begun. OpenAI may keep a copy for up to 30 days for safety reasons, and a temporary chat is not used to improve its models while it stays temporary. If you save it, it becomes a regular chat and follows your normal settings.
Notice what Temporary Chat does not do: it is not an anonymous mode. It is a promise about your history and your memory, not about the provider never seeing the text for a short period.
Does turning memory off change how ChatGPT search works?
It changes less than people expect. Memory controls what ChatGPT knows about you; search controls where it gets facts about the world. A memory-free chat can still browse the web if search is on, and a chat with memory on can still answer from training data without searching. If you are mixing the two settings and getting confusing results, telling whether ChatGPT is actually browsing or guessing is the skill to practise.
What does change is personalization of the search question. With memory on, ChatGPT may fold in details it knows about you, such as your city or dietary habits, when it writes the query or ranks the answer. With memory off, you have to say those details yourself in the prompt, which is more typing but also more predictable. For many people that trade is exactly the point.
What should you use instead of memory?
If you turned memory off because it remembered the wrong things, you may still want stable context. Custom Instructions are a brief you write and control, and nothing is inferred behind your back; custom instructions that actually work shows how to write one. For recurring work with files, a project keeps its context in one place, as explained in ChatGPT Projects versus a regular chat.
A simple rule of thumb: leave memory on for low-stakes preferences like units and tone, turn it off or use Temporary Chat for anything private, and keep a short written brief for the rest.
What to check after you finish
Open the memories list one more time and confirm it is empty or contains only what you want. Start a fresh chat and ask what ChatGPT knows about you. Do this again after any big app update, since OpenAI is still changing how memory works and a setting can move or be renamed. And if your screen does not match anything described here, trust your screen over any article, this one included.
Sources
- Memory FAQ | OpenAI Help Center
- Temporary chat in ChatGPT | OpenAI Help Center
- Memory and new controls for ChatGPT | OpenAI
Checked on November 10, 2026. Features, plans and names change often, so the vendor's own pages are the final word.