How to Use ChatGPT Memory: Manage What It Remembers

ChatGPT Memory makes the assistant feel like it actually knows you. Tell it once that you’re vegetarian, and weeks later it still skips the chicken recipes.

That convenience comes with a few real catches, especially around what sticks when you delete a chat. We tested adding, editing, and wiping memories across the web and mobile apps to map exactly how it behaves.

#What Is ChatGPT Memory and How Does It Work?

Memory lets ChatGPT carry context from one chat into the next, so you stop repeating yourself. According to OpenAI’s Memory FAQ, it works two ways: “saved memories you’ve asked it to remember and chat history, which are insights ChatGPT gathers from past chats to improve future ones.” Those are separate mechanisms with separate controls.

Saved Memories are explicit facts. You tell it “Remember that I work in marketing,” and it files that away. The second type quietly reads patterns across your old conversations, then folds them into future replies. OpenAI’s What is Memory help page describes the two as working together to personalize answers, which is why responses get noticeably sharper the longer you stick with one account and the more it learns about how you actually work.

If you want ChatGPT to follow a fixed style or role in every reply, that’s a different tool. You set that up with custom instructions rather than Memory. Power users layer Memory on top of ChatGPT Projects, which keep separate context for separate jobs.

#How Do You Teach ChatGPT to Remember Something?

The simplest method is to just say it. Start a message with “Remember that…” and ChatGPT saves it. OpenAI’s example is “Remember that I am vegetarian when you recommend recipes.”

It also saves things on its own. If you mention a detail that looks useful later, it may store that as a memory without you asking.

To audit what it has, type “What do you remember about me?” and it lists everything currently saved. In our testing, that prompt surfaced a stored city and job title we’d only mentioned once in passing, which is exactly why the quick check is worth running. It’s the fastest way to catch anything it picked up that you didn’t intend.

#Viewing, Editing, and Deleting Saved Memories

Everything lives in one place: Settings, Personalization, then Manage Memories. Each saved memory shows as its own line, and you can delete them one at a time or clear the whole list.

You don’t even need the menu. OpenAI states that you can simply “tell ChatGPT to delete or forget it,” and the memory goes away conversationally. On the web, the three-dot menu also lets you view history and restore a previous version if you wiped something by accident.

Here’s the part worth bolding: deleting a chat does not delete its memories. OpenAI confirms that saved memories are “stored separately from your chat history,” so a memory survives even after you trash the conversation that created it.

To fully remove a fact, do two things. Delete it in Manage Memories, then delete the original chat too, because either one alone leaves a usable trace.

#Turning Memory Off and Using Temporary Chats

You can switch Memory off anytime under Settings, Personalization, Memory. While it’s off, ChatGPT won’t create or use memories. One catch worth knowing: OpenAI notes that turning saved memory off “won’t delete anything that’s already been remembered,” so flip the toggle and your old list still sits there until you clear it.

For a clean, no-trace conversation, use a Temporary Chat. It won’t reference or update memory at all, which makes it the right pick for anything sensitive. OpenAI states that deleted chat-history data is removed from its systems within 30 days once you switch the setting off. If a whole conversation vanished and you needed it, here’s how to recover missing ChatGPT chat history.

#Memory and Custom GPTs

Not the way you’d expect. As Tom’s Guide reports, custom GPTs you build are effectively stateless, so they don’t carry your saved memories between chats the way the main assistant does.

That distinction matters if you’ve built a custom ChatGPT study helper and wonder why it keeps forgetting you. The fix isn’t Memory; it’s baking the context into the GPT’s own instructions instead.

#The Privacy Catches You Should Know

Memory is convenient, but it’s also a record. Anything you tell ChatGPT to remember sits in your account until you remove it, and it can resurface in later answers when you least expect it. Treat it like a notepad that other prompts can read.

The biggest gotcha is the delete trap covered above, where wiping a chat leaves the memory intact. The second is account scope. Because memories are tied to your login, they sync across your devices, so a memory you added on your phone shows up on your laptop. If you share an account or a device, audit Manage Memories regularly and lean on Temporary Chats for private topics.

#Bottom Line

Treat Memory as a home for short, durable facts, such as your role, your tone preferences, or recurring goals. Don’t use it as a place to paste full templates, which it handles poorly. Audit Manage Memories every few weeks and clear anything stale.

The one real gotcha is the delete trap: wiping a conversation does not erase the memory it created, so remove that separately if you want it gone. For anything sensitive, open a Temporary Chat so nothing is stored at all.

#Frequently Asked Questions