What ChatGPT Remembers About You — and How to Erase It

You open a new chat and ask for help rewriting an email. Before you type a word about the situation, the assistant already suggests a friendly tone and avoids the formal phrasing you complained about last week. Nothing in this conversation told it that. Something from an older chat followed you here.

That something is memory. It is a small, quiet layer of ChatGPT that keeps notes about you between conversations. Most people never open the page where those notes live, which is exactly why it is worth a look.

What actually gets written down

Think of memory as a short list of sticky notes attached to your account. Each note is a sentence or two, not a transcript. If you mention that you work night shifts, a note might say you prefer answers in the evening. If you ask for recipes without cilantro, a note might record that preference. The assistant writes these notes as you chat, and it can update them when you share something new — a new job, a new city, a new project.

The notes are stored with your account, not inside the conversation. That is why a fresh chat can still feel like it knows you. Temporary chats are the exception: they neither read your saved notes nor add new ones, so they stay sealed off from the rest of your history.

Where the notes live and how to clear them

Open Settings, then Personalization, then Memory. You will see a list of the notes ChatGPT has kept. Each one is a plain sentence you can read. Next to each note is a small trash icon. Click it, and that single memory is gone — the assistant will not use it again.

If you would rather start clean, there is a button to delete everything at once. And if you want the feature to stop collecting new notes entirely, you can switch memory off. The setting is available on both free and paid accounts, so you do not need a subscription to take control of it.

• Read first: skim the list of saved notes before deleting anything. Some are useful — a preferred language, a recurring project, a name you use often.
• Delete one at a time: the trash icon beside a note removes only that note. Use it for outdated details like an old address or a joke you no longer find funny.
• Clear everything: the delete-all option wipes the whole list in one step. Good for a fresh start, but it also removes the helpful notes.
• Turn it off: switching memory off stops new notes from being written. You can turn it back on later if you change your mind.

Why old notes can trip you up

Memory does not judge whether a detail is still true. If you mentioned two years ago that you were training for a marathon, that note may still be sitting there, quietly shaping suggestions about running shoes. The assistant has no way to know you stopped. It only knows what was written down.

This is the practical reason to check the list every so often. A quick review catches stale details before they leak into a new conversation and make the assistant sound oddly out of date. It also catches anything you would rather not have stored at all — a health concern, a family situation, a passing remark you did not intend to keep.

Other assistants keep similar notes. Google Gemini has its own memory feature, and Claude uses a project-based context system. The names differ, but the idea is the same: a small store of personal context that travels with you. Knowing where that store lives — and how to open it — is the part that matters.

So the next time a chat feels suspiciously well-informed, you will know where the information came from. A few minutes in the memory settings is enough to see what is there, remove what you no longer want, and decide whether you would rather keep the feature on or off.