ChatGPT Memory prompts to save the useful bits and wipe the rest
Audit what ChatGPT already stored, write remember / forget lines you can stand behind, and turn Memory off when the job is sensitive. Official Memory controls only.
In short: This page contains 15 copy-paste ready prompts, organized into 5 categories with a description and pro tip for each. The first 5 prompts are free instantly, no signup needed. Hand-curated and tested by the AI Academy team.
Audit
3 promptsList what you think you know about me
1/15✨ What it does
ChatGPT lists stored memories in groups and marks keep, edit, or delete. Open Settings, Memory and delete the ones you marked. Do not invent a past.
List the memories you would use about me in this chat. Group: work, personal, preferences, unknown. For each, say if I should keep, edit, or delete. If you have none, say so. Do not invent a biography.
Pro tip: Run this on a new chat. If the list surprises you, Memory is already too loose.
Sensitive leak check
2/15✨ What it does
ChatGPT lists sensitive memory types only. Delete every row that is not none. Re-run the list after you delete.
From anything you remember about me, list items that look like: health, finances, kids, passwords, client names, home address. If none, say none. Do not expand a partial secret into a full one.
Pro tip: Memory is bad at secrets. If it can recite a client, that client is now in the feature. Wipe it.
Stale work facts
3/15✨ What it does
ChatGPT flags memories that contradict [ROLE] and [COMPANY]. Delete the stale ones in settings. Then tell it the new role once.
Here is my current role: [ROLE]. Current company: [COMPANY]. Flag any memory that contradicts this (old employer, old stack, old title). Quote the stale bit. Do not update until I say keep or delete.
Pro tip: Old job memories make ChatGPT write the wrong resume voice. Wipe them the week you switch.
Prompts get you started. Tutorials level you up.
A growing library of 300+ hands-on AI tutorials. New tutorials added every week.
Save on purpose
3 promptsRemember this, tightly
4/15✨ What it does
ChatGPT stores one sentence from [FACT] or refuses a secret. Check the confirmation. If it stored extra color, tell it to forget the extra.
Remember this as one sentence, nothing more: [FACT]. Confirm the exact sentence you stored. If [FACT] contains a secret (key, password, SS number), refuse and tell me to put it in a password manager instead.
Pro tip: One sentence. If you paste a paragraph, Memory will keep the worst clause.
House style, not my life story
5/15✨ What it does
ChatGPT stores each style rule as one sentence and skips your biography. Confirm the list matches [RULES]. Delete extras.
Store only these style rules as memories, each one sentence: [RULES]. Do not store my biography from this chat. Confirm the list.
Pro tip: Memory is decent for I hate em dashes. It is a bad diary.
Team defaults I allow
6/15✨ What it does
ChatGPT writes 3 remember lines with company, product, audience, and not-public. Save those three only. Skip numbers.
I want Memory to know I work at [COMPANY] on [PRODUCT] for [AUDIENCE], and that drafts are not public. Write 3 remember lines. No revenue, no headcount, no customer names.
Pro tip: If a number would change next quarter, do not store it. Pin a file in a Project instead.
Forget
3 promptsForget this fact
7/15✨ What it does
ChatGPT confirms what it forgot about [FACT] and what is still adjacent. If adjacent remains, delete it in Memory settings too.
Forget anything about [FACT]. Confirm what you removed and what you still know that is adjacent. If you still know it, say so. Do not pretend a wipe you cannot do.
Pro tip: Forget this in chat is not always a full wipe. Check the Memory UI.
Wipe a project
8/15✨ What it does
ChatGPT lists memories that mention [PROJECT]. Delete that list in settings. Do not discuss the project again in a Memory-on chat if it was sensitive.
I finished [PROJECT]. List memories that mention it. I will delete them. Do not add new memories from this message.
Pro tip: Finish, wipe, then talk. Otherwise Memory relearns the thing you tried to end.
Turn Memory off script
9/15✨ What it does
ChatGPT writes a 6-line off checklist. Follow it before a sensitive thread. Use a Project with files you control instead of Memory.
Write a 6-line checklist: when to turn ChatGPT Memory off (legal, health, someone else's data, incident), how to disable it, and how to use a Project with files instead.
Pro tip: Off is a switch. Do not rely on please do not remember this as the control.
Like these prompts? There are full tutorials behind them.
Learn the workflows, not just the prompts. 300+ easy-to-follow tutorials inside AI Academy — and growing every week.
Work patterns
3 promptsMemory vs Project vs GPT
10/15✨ What it does
ChatGPT tables Memory vs Project vs GPT for [JOB]. Put prefs in Memory, files in a Project, shared jobs in a GPT. Do not dump files into Memory.
I have [JOB]. Tell me what belongs in Memory (stable prefs), a Project (files), or a Custom GPT (shared job). 9-row table. Use only [JOB] and [CONSTRAINTS].
Pro tip: If it is a file, it is not a memory. People invert this and then wonder why numbers drift.
Weekly memory review
11/15✨ What it does
ChatGPT asks 5 yes/no drop questions and waits. Answer them, then delete in the UI. Do not let it delete on a guess.
Ask me 5 yes/no questions to decide what ChatGPT Memory should drop this week. Then wait. Do not delete until I answer.
Pro tip: A review that waits is useful. A review that auto-wipes is how you lose the one useful pref.
Explain Memory to a teammate
12/15✨ What it does
ChatGPT writes 150 words for [TEAMMATE] on Memory limits. Send it with your policy link. If it marked [VERIFY], add the real policy sentence.
Write 150 words for [TEAMMATE] on what ChatGPT Memory stores, that it can be wrong, and that they should not put client PII in it. From [POLICY] if I pasted one, else label [VERIFY].
Pro tip: Assume they think Memory is a CRM. The note should say it is not.
Incidents
3 promptsWrong memory, correct the record
13/15✨ What it does
ChatGPT forgets [WRONG], stores [TRUE] as one sentence, and confirms. Check Memory settings. If [WRONG] is still there, delete it by hand.
You remember [WRONG]. That is false. The true fact is [TRUE]. Forget the wrong one, store the true one as a single sentence, and confirm both actions.
Pro tip: Corrections in chat can fail. The UI is the source of truth.
Shared-computer warning
14/15✨ What it does
ChatGPT writes an 8-line shared-account warning. Put it on the shared machine, then turn Memory off on that login before anyone else uses it.
Write a 8-line warning for a shared ChatGPT account: Memory will mix people, what to turn off, and why each person needs their own login.
Pro tip: Shared logins plus Memory is how someone else's kid ends up in your work brief.
Export before a wipe
15/15✨ What it does
ChatGPT asks 8 pref questions and builds a post-wipe remember list. Wipe, then paste the list one sentence at a time. Skip health and secrets.
I am about to wipe Memory. Ask me 8 questions to capture prefs I still want, then give me a remember-this list I can paste after the wipe. Do not keep health or secrets.
Pro tip: Export means your notes, not a hidden OpenAI dump. Write the list yourself.
Free tool
Prompt Optimizer
Turn a rough idea into a structured, professional AI prompt.
Frequently Asked Questions
Prompts are the starting line. Tutorials are the finish.
A growing library of 300+ hands-on tutorials on ChatGPT, Claude, Midjourney, and 50+ AI tools. New tutorials added every week.
7-day free trial. Cancel anytime.
Related guides