What should ChatGPT know about me?
The goal isn't maximum disclosure — it's the minimum context that makes answers fit. Here's where that line sits.
When ChatGPT asks what it should know about you, most people either write nothing or write far too much. Both are mistakes. Nothing means generic answers forever. Too much means you've put personal records into a system you don't control, usually without improving a single reply.
The useful test for any piece of information is: would a competent helper need this to give me a better answer? A helper planning your dinners needs to know you cook for three and dislike fish. They don't need your address or your bank.
Worth sharing
The shape of your week
Roughly when you're occupied and when you're free, and what's immovable. Without this, every plan you're given is fiction.
Your household in general terms
How many people you plan for and their broad needs — "two adults and a young child" is enough. Names, birthdays, and schools add nothing.
Food, shopping, and lifestyle preferences
What you eat, what you avoid, how much time you usually have, your general budget level, what you find enjoyable versus a chore.
How you want to be spoken to
Length, formality, whether you want options or a single recommendation, whether you want to be asked clarifying questions. This one line improves nearly every answer.
Practical constraints and tools
Whether you drive, what apps you use, physical limitations that affect plans, how much space you have. Constraints are what make suggestions actionable.
Current priorities
Two or three things you're working toward right now. Keep this current; stale priorities skew suggestions for months.
Keep out of it
- Passwords, PINs, security questions, or any login credentials
- Social Security numbers, passport, driver's licence, or tax identifiers
- Bank, card, or investment account numbers, and full statements
- Your exact home address or precise real-time location
- Medical diagnoses, test results, prescriptions, or therapy content
- Other people's private details shared without their agreement
- Confidential work material your employer's policy doesn't allow
The reasoning is simple: conversations may be retained, may be reviewed by humans in some configurations, and could be exposed in a breach. Assume anything you type could be read by someone else eventually, and share accordingly.
Talking about sensitive topics without disclosing records
You can still get help in sensitive areas by staying general. Instead of pasting a diagnosis, ask what questions to bring to an appointment. Instead of uploading statements, discuss which categories of recurring spending are worth reviewing. AI is not a substitute for a doctor, lawyer, accountant, or therapist, and shouldn't be treated as one.
Review what's saved
Check your assistant's memory or saved-info screen every couple of months and delete what's no longer true. Outdated context is a quiet cause of unhelpful answers. Our guide to making ChatGPT remember your preferences covers how to manage it.
Not sure what context is worth writing down for your situation? The free AI assessment suggests the specific categories that matter for you.
If you'd rather answer guided questions than stare at a blank box, our Personal AI Profile builder writes the whole thing for you and shows you what to leave out.
Not sure where AI fits your life?
The free AI assessment takes a few minutes, needs no account, and gives you a personalized list of everyday places AI could help.
Start My AI Setup