Hidden Features

Memory

Persistent facts ChatGPT remembers across chats.

ChatGPT Memory

What Memory does

With Memory on, ChatGPT can retain selected facts about you across conversations — your role, ongoing projects, preferences — so a new chat is not a blank slate. Used well, it removes the friction of re-explaining yourself and makes answers steadily more tailored over time. It is different from custom instructions: instructions are what you set deliberately, while Memory accumulates as you talk, picking up details it judges worth keeping. That automatic quality is both its strength and the thing to keep an eye on.

Curate it, don't just let it grow

Memory helps only if it stays accurate. It can remember something that was true once, a passing detail you did not want kept, or a wrong inference — and then quietly colour later answers. Review what it has stored periodically, delete anything stale or mistaken, and correct it when it remembers wrongly. A curated memory is an asset; an unmanaged one gradually introduces noise. Treat it like a shared notebook you tidy now and then, not a box you never open.

Frequently asked
How does ChatGPT Memory work?

With Memory enabled, ChatGPT retains selected facts across conversations — your role, projects, and preferences — so new chats start with useful context instead of blank. It accumulates as you talk.

Is ChatGPT Memory the same as custom instructions?

No. Custom instructions are what you set deliberately; Memory accumulates automatically from your conversations. They complement each other — instructions for standing preferences, Memory for evolving context.

Should I manage what ChatGPT remembers?

Yes. Review it periodically and delete anything stale, wrong, or that you did not want kept, since old or mistaken memories can quietly skew later answers. A curated memory is an asset.