Manage What ChatGPT Remembers About You

Take Control of Your Memory

ChatGPT’s memory is only useful if it’s accurate. Outdated information, finished projects, or changed preferences left in place will skew responses without you knowing why. This lesson shows you where the settings are, how to correct them, how to delete them for good, and how to disable what needs disabling.

Access Your Memory

Go to Settings > Personalization > Memory. Depending on your memory experience, you’ll find some or all of these settings: Memory, the master switch; Reference saved memories, for stored memories; Reference chat history, for leveraging past conversations; the memory summary; and Manage or Saved memories, for the list of memories. If a setting described here doesn’t appear, it isn’t available for your account yet. You can also simply ask ChatGPT what it remembers about you.

Correct Rather Than Endure

The memory summary gives an overview of what ChatGPT can use to personalize its responses. When available, you can enter a correction there, highlight a passage to fix it, or choose Don’t mention this again to stop an item from coming up. Note the nuance: this last option changes ChatGPT’s behavior but doesn’t delete the source, whether that’s a conversation, a file, or an email. You can also correct things in conversation: “What you know about my job is outdated; I’m now a Sales Director.”

Delete for Good

Deleting information means removing it from wherever it’s stored, and it can be stored in several places: the memory summary, saved memories, regular and archived conversations, files in the Library or Space, and connected apps. The most surprising point: deleting a conversation doesn’t necessarily delete the memory derived from it, because saved memories are stored separately. To erase a memorized item, delete both the memory and the conversation where you shared it. Conversely, deleting a memory removes it from future personalization but doesn’t erase its mention in past conversations. Updates can take a little time, and OpenAI may retain logs of deleted memories for up to 30 days for security and debugging.

Disable, Knowing the Effects

Turning off Memory doesn’t erase your conversations; if you reactivate it later, ChatGPT can again retain information from conversations remaining in your history. Turning off Reference chat history schedules the deletion, within 30 days, of what ChatGPT had drawn from your past conversations, without touching those conversations or your saved memories. Turning off Reference saved memories also cuts off reference to the history. Finally, the Delete and turn off memory option in the summary erases what the summary contains and disables memory, without deleting your conversations.

For a single sensitive exchange, there’s no need to disable everything: open a temporary chat, choosing Unpersonalized if you also don’t want it to draw on what ChatGPT already knows about you.

Management Strategies

The first strategy is a regular check-in. Once a month, review the summary or list of memories, delete what’s no longer relevant (old projects, outdated info), verify that no sensitive data has slipped in, then add important new items. Ten minutes is enough, and you’ll feel the difference immediately.

The second is structured memory: instead of a jumble of information, give ChatGPT context organized by domain.

Remember for my professional context:
- Industry: EdTech
- Role: Product Manager
- Tools: React, Node.js, PostgreSQL
Remember for my personal context:
- I live in Lyon
- I cook often, recipes under 30 minutes

The third separates universes: for a folder that needs to stay siloed, a project can have its own memory, limited to its conversations (Lesson 9).

Costly Mistakes

Never checking your memory is the most insidious mistake: outdated information skews responses, and you’ll wrongly blame the tool. Storing sensitive data is the most dangerous: never memorize passwords, access codes, or confidential data. Believing that deleting a conversation takes its memories with it is the most common. And disabling memory on principle deprives you of a real quality lever: reserve discretion for exchanges that justify it, using temporary chat.

Your Turn

Your monthly check-in. Go to Settings > Personalization > Memory and review what ChatGPT remembers about you. Correct an inaccurate item, choose Don’t mention this again on another if the option exists, and delete an unnecessary memory. Then add three useful professional items, structuring them as above. Finally, open an unpersonalized temporary chat and verify that ChatGPT doesn’t refer there to what it knows about you.

Official sources: Memory in ChatGPT (OpenAI Help Center: settings, summary, corrections, deletion, disabling, 30-day retention) · Temporary chat in ChatGPT (Personalized and Unpersonalized) · Projects in ChatGPT (project memory)

Key Takeaways

• Settings are in Settings > Personalization > Memory; they vary by account
• Correct in the summary or in conversation; “Don’t mention this again” doesn’t delete the source
• Deleting a conversation doesn’t necessarily delete the memory derived from it: delete both
• Turning off Reference chat history erases within 30 days what was drawn from it, without touching conversations
• For a sensitive exchange, prefer an unpersonalized temporary chat