What ChatGPT, Claude, and Gemini Remember About You (and How to Control It)

Every major AI assistant now keeps notes on you. ChatGPT, Claude, and Gemini all ship with memory switched on for personal accounts, and all three learn from your conversations whether or not you ever open the settings panel.

Most people never do. So the questions they eventually type are the obvious ones. What does ChatGPT remember about me? How do I delete it? Does a temporary chat actually disappear?

Here are the answers, tool by tool, with the menu paths checked against each vendor's help pages in October 2026. Fair warning: these menus move. Claude and Gemini have both renamed theirs in the past few months.

What do ChatGPT, Claude, and Gemini remember about you?

All three remember your preferences, projects, and recurring details from past chats, and all three let you view, edit, pause, and delete that memory. ChatGPT keeps saved memories and can reference your chat history. Claude keeps editable memory entries, with a separate space per project. Gemini draws on your past chats while Keep Activity is on. Deleting a chat doesn't delete what was learned from it.

	ChatGPT	Claude	Gemini
Where to look	Settings → Personalization → Memory	Settings → Your Claude → Manage (next to Your memory)	Settings & help → Personal Intelligence
Turn it off	Toggle saved memories and chat history separately	Pause (keeps it) or Reset (deletes it all)	Turn Memory off
Private chat	Temporary Chat	Incognito (ghost icon)	Temporary Chat
Private chats kept for	Up to 30 days, for safety	Not saved to history	Up to 72 hours
Training switch	Data controls → Improve the model for everyone	Privacy → model-improvement toggle	Tied to Keep Activity

That table covers what to click. The rest of this post covers what those clicks actually do, because a few of them don't do what you'd expect. If you want the deeper comparison of how each memory works, we covered that separately.

How do you see and delete what ChatGPT remembers?

Open Settings, then Personalization, then Memory. OpenAI's help page says the exact controls vary by account, but you'll see some mix of Reference saved memories, Reference chat history, and a Manage option.

The thing is, those are two different systems. Saved memories are the notes ChatGPT has written down, either because you asked ("remember that I write in British English") or because it decided something was useful. Chat history referencing draws on your past conversations more broadly, and it has no list for you to read.

So the Manage panel feels like a full inventory. It isn't. It's the part you can see.

You can delete saved memories one at a time or clear the lot. If you switch off chat history referencing, OpenAI says what it remembered from past chats is scheduled for deletion within 30 days. Deleted saved memories may also sit in OpenAI's logs for up to 30 days for safety and debugging.

How do you see and delete what Claude remembers?

Go to Settings, then Your Claude, and select Manage next to Your memory. If your app still shows the older layout, it's under Settings, then Memory. Anthropic's help center lists both.

Claude's memory is the easiest of the three to correct. It's readable text, so a wrong detail is an edit, not a wipe. Each project also keeps its own memory space, which keeps client work from leaking into your side project.

When you want to stop it, you get two choices, and they're very different:

• Pause memory. Claude keeps what it has but stops using it and stops adding to it.
• Reset memory. Everything is permanently deleted, project memories included. Anthropic is blunt that this can't be undone.

There's a third switch worth knowing. Search and reference chats lets Claude look through your past conversations, and you can turn it off on its own. Claude also doesn't automatically save sensitive topics like health, religion, or politics unless you switch on Include sensitive topics in memory.

How do you see and delete what Gemini remembers?

On gemini.google.com, open Settings & help, then Personal Intelligence, and turn Memory on or off. That same area holds Instructions for Gemini, the standing preferences you've given it, which you can edit or delete one by one.

Two catches from Google's help pages. First, memory of past chats only works while Keep Activity is on. Second, the conversations themselves live in Gemini Apps Activity, and that's where you delete them or set them to auto-delete. Google notes that after you delete a chat, there may be a short delay before Gemini stops using it to personalize your answers.

Availability is narrower than the other two. Google says these features need a personal Google Account, so work, school, and supervised accounts don't get them, and some memory features are currently US-only.

Does deleting a chat delete the memory?

No. This is the one that catches almost everyone.

Anthropic says it plainly: when you delete a conversation, the memory entries generated from it aren't removed. ChatGPT works the same way in practice, since saved memories live in their own panel. The chat goes. What the tool learned from it stays.

Picture someone who mentions a health scare in one conversation, regrets it, and deletes the chat. If the tool already saved that detail, it's still there, quietly shaping answers months later. To remove something for real, delete the chat and then delete the memory.

Is turning off memory the same as opting out of training?

Also no. They're separate switches that do separate jobs.

Memory decides what the tool uses to personalize your answers. Training decides whether your conversations help improve future models for everyone.

• ChatGPT: Settings, then Data controls, then Improve the model for everyone. OpenAI says it's on by default for Free, Plus, and Pro, and you can turn it off without losing your chat history.
• Claude: Settings, then Privacy, then the toggle to help improve the models. Anthropic's privacy center says chats are kept for five years with it on and 30 days with it off.
• Gemini: Google ties this to Keep Activity. With it off, chats are kept for 72 hours to respond to you and keep the service safe.

You can turn off memory and leave training on, or the reverse. Decide each one on purpose.

Do temporary and incognito chats really forget?

Mostly. They don't show up in your history, they don't create memories, and they aren't used for training. ChatGPT's Temporary Chat, Claude's incognito mode (the ghost icon, top right), and Gemini's Temporary Chat all promise that much.

But "not saved" doesn't mean "never stored." OpenAI may keep a temporary chat for up to 30 days for safety. Google keeps Gemini Temporary Chats for up to 72 hours. And in ChatGPT, if you save a temporary chat, it becomes a regular one and follows your normal settings.

Use them for the one-off question you don't want shaping anything. Don't treat them as a shredder.

The context you actually control

Here's the pattern underneath all of this. Every one of these memories is written by the tool, stored by the tool, and governed by menus that change every few months. You can audit them. You can't really own them.

And what they keep is thin. A preference here, a project name there. None of it follows you to the next tool, and none of it holds the parts of you that make an answer sound like yours.

So flip the arrangement. Keep the context that matters in a plain document you own, paste it in when you want it, and let built-in memory handle the small conveniences. Then deleting a memory panel stops feeling like losing something, because it was never the real copy.

The easiest place to start is values: what you care about, what you won't do, and how you make decisions. The free Personal Constitution builder takes about thirty minutes, and it stays yours no matter which settings menu moves next.