How to Make ChatGPT Remember You

You've told ChatGPT who you are more times than you can count. That you write for a living. That you've got a teenager, a dog, and a move on the horizon. That you want short sentences and no corporate filler. And still, every new chat opens like you've never met, and you're back at the start, explaining yourself to a stranger who greeted you warmly yesterday.

So you do the reasonable thing. You go looking for how to make the thing remember you.

Good news: it can. Better than most people realize, actually. ChatGPT holds onto who you are in three different ways, and they stack. Let's turn each one on, in the order that matters, and then talk about the wall they all eventually hit.

Does ChatGPT actually remember you?

Yes, with the right settings on. ChatGPT has three separate ways of carrying your context from one conversation to the next: automatic memory, custom instructions, and projects. Most people use one of them by accident and never touch the other two, which is why the remembering feels so hit-or-miss. Set up all three on purpose and the experience changes.

Here's how to do that.

Step 1: Turn on memory and feed it

ChatGPT's memory is the automatic layer. Open Settings, go to Personalization, and make sure Memory is switched on. Once it is, ChatGPT quietly notes things across your chats — that you're writing a book, that you prefer Oxford commas, that you live abroad — and pulls them forward when it thinks they're relevant.

You don't have to leave it all to chance, though. Any time something matters, say so directly: "Remember that I run a small agency and write all our client copy myself." It files that away on the spot. And when the memory gets cluttered or wrong, go back to that same Personalization panel and prune it. The list of everything it's saved is right there, editable.

Memory is the easy win. It's also the layer you steer the least, so don't lean your whole identity on it.

Step 2: Set your custom instructions

This is the layer you actually control, and it's the one most people skip. In Settings, under Personalization, open Custom Instructions. You'll find two boxes: one for what you want ChatGPT to know about you, and one for how you want it to respond.

Treat the first box like a standing introduction. Your role, your work, the things you're always coming back to. Treat the second like a style guide. "Skip the preamble. Short paragraphs. Don't tell me a topic is important, just get into it." Whatever you write here rides along on every single chat, no reminding required.

The difference is immediate. Memory is ChatGPT's running notes on you. Custom instructions are your notes to ChatGPT. One is a guess; the other is an instruction. If you want a deeper version of this — setting it up across ChatGPT, Claude, and Gemini at once — there's a fuller walkthrough on that.

Step 3: Use a project for the big stuff

For anything with real weight — a book, a business, a long-running side thing — a single set of instructions isn't enough. That's what projects are for. Spin one up, give it standing instructions and a file or two, and every chat inside that project starts from the same context. Your novel project knows your characters. Your client project knows the brand.

It's the most powerful of the three, because now you're handing ChatGPT a real briefing instead of hoping it remembers. Do this for the two or three things you work on most and you'll stop re-explaining them for good.

Three layers, switched on, working together. For most of what you do inside ChatGPT, that's genuinely most of the way there.

Where this stops working

Here's the part the setup guides leave out.

Open Claude to draft something, and you're a stranger again. Try Gemini for a research pass, same thing. Every layer you just built — the memory, the instructions, the projects — lives inside ChatGPT and nowhere else. You did the work once, and the work stayed behind the moment you switched tabs.

And even inside ChatGPT, what it holds is thin. Memory captures preferences and stray facts, the surface stuff. It doesn't hold the things that actually make your work sound like yours: the way you build an argument, the stories you reach for, the principles you won't bend on. The real you is a lot more than a list of settings the model happened to save. If you want the longer version of why that gap exists, ChatGPT memory, Claude Projects, and why your AI still forgets you walks through it.

So you can do everything right and still hit the same two walls. The picture is shallow, and it's trapped in one app.

The version that travels

There's a way to give your AI context that doesn't live inside any single tool. You write the durable facts down once — who you are, how you sound, what you're building, who's in your orbit — in a plain document you own, and you carry it to whatever model you're using that day. A context anchor, it's called, and it's the same idea as custom instructions, just portable.

Paste it into a ChatGPT chat. Drop it into a project. Wire it into Claude or Gemini or next year's tool that doesn't exist yet. The tool changes; the context doesn't. You stop rebuilding yourself inside every new app and start showing up the same in all of them. ChatGPT's own features get sharper too, because now you're feeding them a real account of yourself instead of letting the model guess.

Let the tools keep their convenient little memories. The full picture of you should belong to you.

Start with one

Go back to that moment ChatGPT greeted you like it knew you, then evaporated the second you opened a new chat. Three settings switched on will close most of that gap inside ChatGPT. What closes it everywhere is a picture of yourself that doesn't live inside any single tool, so it's there every time you open a new one.

If you want the bigger story on how this works, start here. If you'd rather just build the thing, write your first anchor. It's the free one, you can finish it this afternoon, and the next tool you open won't have to meet you cold.