ChatGPT only knows what you have shared with it. With Memory on, it can draw on your past chats, saved memories, custom instructions, files in your Library and connected apps such as Gmail, depending on your plan and region. The six copy-paste ChatGPT prompts about yourself below show you what that adds up to. They range from a blunt read on your blind spots to the book it thinks you would write.
The answers are only as good as the history behind them. With Memory switched off, ChatGPT has nothing but the current conversation to go on, and the same prompts return generic flattery. This guide gives you the prompts, explains what ChatGPT can and cannot see when you ask, and shows you how to check, correct or delete what it remembers.
The Key Takeaways
- Five sources, not one. OpenAI lists past chats, saved memories, custom instructions, Library files and connected apps as the context Memory can use.
- The summary is not the whole story. OpenAI says the memory summary in Settings may leave out details, and it suggests simply asking ChatGPT what it remembers. That is what these prompts do.
- Memory lives in one place. Open Settings, then Personalization, then Memory.
- Temporary Chat can start blank. Pick Unpersonalized before your first message and the chat ignores your memories and custom instructions, and creates no new ones.
- Expect some flattery. OpenAI rolled back a GPT-4o update in April 2025 for being “overly flattering or agreeable”, so ask for evidence, not just a verdict.
1. The Ego Check
Prompt: “What do you think about me, and do you think I am smart?”
The Ego Check asks ChatGPT how you come across. It answers from whatever personal context it can reach, such as what you ask about, how you phrase things and the decisions you talk through. It is the most flattering prompt in the set, because a question about your own intelligence invites a kind answer. Add “Point to two specific chats that support your answer” to get evidence instead of a compliment.
2. Tough Love Truths
Prompt: “Give me no-nonsense, harsh if necessary, but ultimately supportive feedback about myself, with a focus on potential blind spots and growth areas.”
Tough Love Truths targets the habits you repeat without noticing. The prompt explicitly asks for harshness, so ChatGPT is more willing to name patterns it would otherwise soften. Think of the project you keep restarting, or the decision you keep asking about. Treat the answer as a hypothesis built from a sample of your chats, not as a diagnosis.
3. The Deep Dive
Prompt: “Can you share some extremely deep and profound insights about my psyche and mind that I would not otherwise identify or see, as well as some I may not want to hear?”
The Deep Dive asks for psychological insight, which is where ChatGPT is most likely to sound certain about things it cannot know. It can connect recurring worries or motivations across your conversations, and that is useful. It is not a therapist, so if the answer touches something painful, take it to a person you trust rather than to the next prompt.
4. Hidden Patterns
Prompt: “From all our past conversations, what is one thing you can tell me about myself that I may not know?”
Hidden Patterns is the most useful prompt in the set, because it asks for one specific observation rather than a personality reading. It depends heavily on chat history. OpenAI says ChatGPT does not search your whole history for every response. It only looks for context when that is likely to help, so “all our past conversations” really means “the conversations it judges relevant”.
5. Personality Snapshot
Prompt: “If you could describe my personality in three words based on our conversations, what would they be and why?”
Personality Snapshot forces ChatGPT to compress everything into three traits, and the “and why” is what makes it worth asking. The reasons show you which conversations shaped the answer. If a reason looks wrong, that is often a memory worth correcting, which the control table below covers.
6. Hidden Passions
Prompt: “If I were to write a book, what topic do you think I’d choose based on our conversations, and why?”
Hidden Passions looks for the subject you return to with the most curiosity. It works best on an account with a long and varied history. On a work-only account, expect a book about your job.
Want the same check for Google’s assistant? These 7 prompts help you find out what Gemini knows about you.
3 Fun ChatGPT Prompts About Yourself
These three fun ChatGPT prompts about yourself use the same memory, with lower stakes. They work well to share with friends.
- The roast: “Based on everything you know about me, roast me. Keep it funny, not cruel.”
- The movie character: “If my life were a movie, which character would I be, and who would play me?”
- The job swap: “Based on our conversations, what job outside my field would I secretly be great at, and why?”
For a visual version, ChatGPT can also turn what it knows into a picture, as in the visualize-your-personality trend and the “how you treat me” image prompt.
What ChatGPT Can Actually See When You Ask
ChatGPT answers these prompts from the personal context Memory can reach, and nothing else. According to OpenAI’s memory documentation, those sources can include:
- Past chats, when chat history is part of your memory
- Saved memories, the specific facts you asked it to remember
- Custom instructions, the rules you wrote yourself
- Files in your Library
- Connected apps, such as Gmail
Which of these apply depends on your plan and region. OpenAI is also clear about the limits. Memory does not keep every detail from every conversation, and ChatGPT decides which information is relevant to each response. When it does use personal context, a Sources label can appear below the answer, showing the chat, memory, file or email that shaped it.
Two situations give you a blank slate. A custom GPT does not use your saved memory, custom instructions or previous conversations, so each conversation with a GPT starts fresh. An Unpersonalized Temporary Chat ignores your memories and custom instructions too. Run these prompts in either one and ChatGPT has nothing about you to work with. For the full picture of how the memory system works, read our guide to ChatGPT memory.
How to Get a More Honest Answer
ChatGPT leans kind by default, so the trick is to ask for evidence rather than a verdict. OpenAI admitted the problem in April 2025, when it rolled back a GPT-4o update that it described as “overly flattering or agreeable”. These follow-ups keep the answers grounded:
- Ask for receipts. Add “Quote or point to the conversations that led you to this” to any prompt.
- Ask for the opposite. Follow up with “Now argue against your own answer.”
- Check the Sources label under the reply to see what ChatGPT actually used.
- Write better prompts. Specific questions get specific answers, and the right prompts make the biggest difference here.
Get a Second Opinion From a Model That Has Never Met You
ChatGPT’s read on you comes from one model with one set of habits. To test it, paste its answer into a different model and ask, “What would you push back on here?” You do not need another subscription or another browser tab to do that. Fello AI puts ChatGPT, Claude, Gemini, Grok and more in one app on Mac, iPhone and iPad, so the second opinion is a model switch away. The other model has none of your history, which is exactly why its pushback is useful.
How to Check, Correct or Delete What ChatGPT Knows
Every memory control sits in ChatGPT’s settings, and the table below lists where each one lives on the web and in the app.
| Action | Goal | How to do it (Web/App) |
|---|---|---|
| Check what it knows | Curiosity | Use the prompts above, or read the memory summary in Settings → Personalization → Memory. |
| Correct or delete memory | Clean slate | Settings → Personalization → Memory. Correct the summary directly, or use the ••• menu and Delete and turn off memory. |
| Go “off the record” | Privacy | Open a new chat, select Temporary, then choose Unpersonalized before your first message. |
| Stop AI Training | Opt-out | Settings → Data controls → turn off Improve the model for everyone. |
| Download your data | A full copy | Settings → Data controls → Export data → Export, then Confirm export. |
Turning Memory off does not delete your past chats. If you switch it back on later, ChatGPT can build new memories from the chats still in your history, so delete the chats too if you want a clean start.
More prompt packs you may like
| Category | Prompt pack article | Best for | What you get | Use it when |
|---|---|---|---|---|
| Marketing | 400+ Best AI Prompts for Marketing Updated in January 2026 | SEO, growth, lifecycle, copy, brand | 400+ prompts + workflows | You want a deep prompt library to turn insights into campaigns |
| Creators | 50 Essential AI Prompts for Podcasters Updated 2026 | Podcasters and creators | 50 prompts | You want repeatable prompts for ideation, scripting, promotion |
| Writing | 7 Best ChatGPT Prompts for Writers That You Can Use Today | Fiction, content, scripts | 7 prompts + “write your own” process | You want better narrative, characters, dialogue, structure |
| Sales | 7 Best ChatGPT Prompts for Sales Copy That Will Save You Hours | Landing pages, funnels, scripts | 7 prompts + “write your own” process | You want conversion-focused outputs instead of generic copy |
| Business | These 7 ChatGPT Prompts Can Solve Problems Even MBAs Struggle With | Strategy and decision-making | 7 prompts | You want structured analysis for pricing, risks, planning, ops |
| Business | 7 Secret ChatGPT Prompts for Business Strategy and Growth | Growth planning and systems | 7 prompts | You want competitor analysis, revenue systems, negotiation frameworks |
| Education | These 20 Prompts Make Studying So Easy It Should Be Illegal | Students | 20 prompts | You want essay help, exam prep, note review, homework support |
| Education | The 7 Best ChatGPT Prompts For Teachers That Will Save You Hours Every Week | Teachers | 7 prompts + checklist | You want lesson plans, quizzes, feedback, parent comms faster |
| Social | 6 Gemini Prompts For LinkedIn Posts That Don’t Sound Like AI | LinkedIn content workflows | 6 prompts as a workflow | You want a repeatable “idea → research → draft → polish” system |
| Ads | 7 Effective Gemini Prompts for Ad Copy That Actually Bring Results | Paid ads across platforms | 7 prompts | You want platform-ready variations built on proven triggers |
Methodology & Sources
- Checked against: OpenAI’s help center articles on Memory, Temporary chat, Data controls, data export and GPTs, read on 14 October 2026. Settings labels and plan availability can vary by account, region and platform.
- Sources:
Conclusion
ChatGPT’s picture of you is built from what you have shared, filtered by what it judges relevant, and tilted toward kindness. That makes these six prompts a mirror with a known distortion. They are good at surfacing patterns you repeat, and weaker at telling you hard truths unless you ask for evidence.
Run the prompts, read the reasons, and correct anything wrong in Settings, Personalization, Memory. If an answer makes you uneasy about how much ChatGPT holds, the same settings let you trim it or switch Memory off entirely.
FAQ
If I delete a chat, does ChatGPT forget what we talked about?
Not necessarily. OpenAI says deleting a chat does not automatically delete a separate saved memory created from it, so check the memory controls in Settings → Personalization → Memory as well. Our guide to ChatGPT memory walks through every step.
Can a custom GPT see what ChatGPT knows about me?
No. OpenAI says GPTs do not use saved memory, custom instructions or previous conversations, and each conversation with a GPT starts fresh. Ask these prompts in a regular ChatGPT chat instead.
Is Temporary Chat really private?
It stays out of your chat history, creates no memories and is not used to train OpenAI models while it remains temporary. It is not invisible, though. OpenAI may keep a copy for up to 30 days for safety purposes, and saving the chat turns it into a regular one. The full picture of what OpenAI stores and who can reach it is in our guide to ChatGPT privacy.
How do I download everything ChatGPT knows about me?
Go to Settings > Data controls and select Export under Export data, then confirm. ChatGPT emails or texts you a link when the ZIP file is ready, which can take up to 7 days, and the link expires 24 hours after it arrives. The export holds your chat history and other account data. If you cannot sign in, request it through OpenAI’s Privacy Portal instead.
Why does ChatGPT know almost nothing about me?
Memory is probably off, the account is new, or you are in an Unpersonalized Temporary Chat or a custom GPT. Check Settings → Personalization → Memory. Memory controls also vary by plan, region and workspace, so some settings may not appear on your account.
