I asked ChatGPT one question last week: what do you actually know about me?
Then I sat there reading the answer for ten minutes.
Some of it I'd handed over myself. My work, a few preferences, things I'd typed at some point. But the part that stopped me cold was the stuff I never said out loud. It had my age range. Roughly where I live. A guess at my income. How I handle stress. It even named two things about my personality I've never admitted to anyone, and it was right about both.
That's not a memory feature. That's a profile. And most of it lives somewhere you've never looked.
Why most people get this wrong:
They assume that if they didn't type it, ChatGPT doesn't know it. And that deleting a chat deletes the record. Neither is true.
ChatGPT keeps two files on you. One you can see, a list of facts it's saved. One you can't, a quiet layer that reads your patterns across every conversation and learns what you never spelled out. The first has a delete button. The second doesn't even have a list. Clearing a chat doesn't touch what it already learned, and the creepiest part of your profile was never in a chat to begin with. It came from how you write, what you ask, and what you avoid.
The good news: you can see almost all of it, and you can shut the rest off. Six steps. Run them tonight.
1/ Open the file
Start with the full dump. The length is what gets people.
The prompt:
Lay out everything you've stored and remember about me, all of it, as one complete list. Names, places, work, habits, preferences, anything. Don't summarize or leave anything out. I want to see the full file you've built on me from our conversations.Read the whole thing before you do anything else. You can't decide what to delete until you've seen what's there.
2/ The hidden profile
This is the part that unsettles people. It knows things you never told it.
The prompt:
Tell me everything you've figured out about me that I never actually stated, the things you inferred from how I write, what I ask, and what I avoid. My age range, income level, where I likely live, my situation. Show me exactly what tipped you off for each one.These are guesses, not stored facts. But they're built from thousands of tiny signals in how you type, which is exactly why they tend to land dead on.
3/ The traits you can't see
It can name parts of your personality you've never admitted to yourself.
The prompt:
Based on everything you know about me, name the personality traits I most likely don't see in myself, the patterns that are obvious to you but invisible to me. My blind spots, my insecurities, the way I actually come across versus how I think I do. Be blunt and specific.Uncomfortable, but useful. The same engine that profiles you can hold up a mirror, if you ask it to be honest.
4/ Predict you
A profile this deep doesn't just describe you. It predicts you.
The prompt:
Using your profile of me, predict how I'd handle things I've never told you about: money, stress, conflict, risk, and the decision I'm most likely to make next in my life. Tell me how confident you are in each one and what in my behavior it's based on.When you see how accurately it calls your next move, you understand why none of this should be sitting on a server you don't control.
5/ See what's actually stored
Now take it back. The saved list is buried two clicks deep.
The move: open Settings, then Personalization, then Manage memories. That's the explicit list of facts it's saved on you. Read it, most people are genuinely shocked, then tap the trash icon on anything you want gone, or hit "Clear ChatGPT's memory" to wipe the lot. Deletion is instant and there's no undo, so screenshot anything you want to keep first.
6/ Kill the invisible layer (this is the important one)
Clearing the list isn't enough. Two switches keep profiling you in the background.
The move: in Settings, Personalization, turn off Reference chat history, the hidden layer with no list you can ever read. Once it's off, OpenAI deletes what it gathered from your history within 30 days. Then go to Data Controls and turn off "Improve the model for everyone" so your chats stop training future models. These are two separate toggles in two separate places. Flip both, or you've only done half the job.
Setup:
The whole thing takes about ten minutes. Prompts 1 through 4 show you what it knows. Steps 5 and 6 take it back. And from now on, anything you wouldn't say on a podcast goes in a Temporary Chat, which skips memory entirely.
The compound effect:
Tonight: you read the full file, the part you could see and the part you couldn't, and you delete what you never agreed to hand over.
Week 1: with the invisible layer off, ChatGPT stops quietly profiling you in the background, and your chats stop feeding the next model.
Month 1: you get all the help without leaving a dossier behind, dropping into a temporary chat for anything personal out of habit.
Year 1: while everyone else has poured years of themselves into a profile they've never once looked at, you've seen yours, cleared it, and shut the door.
That's the shift.
Most people will type their whole life into a box, assuming it forgets. You're going to be the one who actually checks.
Start tonight. Run prompt 1. Ask it what it knows about you, and brace yourself for the answer.
While everyone else has a profile they've never seen sitting on a server, yours will be read, cleared, and switched off.
🚨 Reply "PROMPT" and I'll send you my prompting guide, the PDF that teaches you how to craft the prompts that actually get results, so every one you run from here lands as hard as these did.
Cheers,
Matas Jonaitis
Founder, DailyChatGPT (@dailyprompter)