If you've been using ChatGPT for more than a few weeks, it knows things about you. Your name, your job, your preferences, maybe even your kids' names or your favorite programming language. Some of it you told it to remember. Some of it, it figured out on its own.
Most people have no idea how much is in there. I didn't either, until I actually looked.
ChatGPT has two memory systems
This is the part that surprises most people. ChatGPT doesn't just have one place where it stores things about you. It has two completely separate systems.
Saved Memories are the ones you explicitly asked it to remember. Things like "Remember that I prefer Python over JavaScript" or "My dog's name is Max." You can see these by going to Settings > Personalization > Manage Memories. They show up as a clean list.
Chat History inferences are the sneaky ones. These are things ChatGPT has picked up from your conversations without you asking. Maybe it noticed you always ask for concise answers, or that you work in marketing, or that you tend to ask about React. These don't show up in the Manage Memories screen. They live in the background and quietly shape how ChatGPT responds to you.
When most people check their memory, they only look at the first type and think, "Oh, there's only 12 things in here." Meanwhile, the second type might have dozens of inferred preferences you've never seen.
How to check what it knows
The Manage Memories page only shows your saved memories. To see the full picture, including all the stuff ChatGPT inferred on its own, you need to ask it directly.
Open any ChatGPT conversation and type something like:
The response will probably be longer than you expect. You'll see things you forgot you told it, things you never explicitly told it, and maybe a few things that are just wrong (it happens).
The export trick most people miss
If you want a clean, copy-friendly version of everything ChatGPT knows, there's a better prompt that formats the whole thing as a code block. We put together a full export prompt that's specifically designed to pull out every category of stored memory: instructions, personal details, projects, tools, corrections, all of it.
The key thing about the export prompt is that it asks ChatGPT to preserve your exact words where possible. If you told it "never use bullet points in your responses," you want that exact instruction carried over, not a paraphrased version.
Once you have the export, you can review it, clean up anything outdated, and then do whatever you want with it. Paste it into another AI tool, keep it as a personal backup, or just use it to understand what you've been teaching ChatGPT over time.
What to do with your exported memory
The most obvious use is moving it somewhere else. If you've been thinking about trying Claude or Gemini, you don't have to start from scratch. Your preferences, your style instructions, your personal context: all of that can come with you.
We have step-by-step guides for the most common migrations:
ChatGPT to Claude takes about 2 minutes. Claude has a dedicated import flow where you just paste your export and it handles the rest.
ChatGPT to Gemini is a little different since Gemini doesn't have a bulk import feature yet. But you can paste your preferences through Gemini's Settings or drop the whole thing into a conversation and ask it to remember everything. Still just a couple of minutes.
Even if you're not planning to switch, it's worth exporting your memory once just to see what's in there. Think of it as an AI audit.
The bigger picture
The fact that this is hard to do says something about where AI tools are right now. Every platform wants to learn about you, but none of them make it easy to take that knowledge with you when you leave. Your preferences are basically trapped.
That's the whole reason we built ImportMemory. You spent months teaching your AI how you like things done. That effort belongs to you, not to whichever tool you happened to start with.
Go check what ChatGPT knows. You might be surprised.