What ChatGPT Memory Does
Memory gives ChatGPT a persistent understanding of who you are across conversations. Without it, every chat starts from scratch — you are a stranger every time. With memory on, ChatGPT carries relevant facts forward automatically.
What gets remembered:
- Your job title and industry
- Preferred response format (bullet points, short answers, code blocks)
- Ongoing projects and their tech stack or context
- Personal preferences ("I prefer metric units," "I am vegetarian")
- Communication style preferences ("be direct," "skip the disclaimers")
- Explicit instructions you give ("always format dates as DD/MM/YYYY")
What does not get remembered:
- The full text of past conversations
- Sensitive data like passwords, card numbers, or health details (ChatGPT actively avoids storing these)
- One-off context that is only relevant to a single session
How to Enable Memory
- Click your name or avatar in the bottom-left sidebar
- Go to Settings → Personalization
- Toggle Memory to on
- Start a new conversation — memory is now active
Memory begins saving automatically. You will not see a notification each time something is saved; check Manage memories periodically to see what has accumulated.
How to View and Manage Saved Memories
- Settings → Personalization → Manage memories
- Scroll through the list of saved facts
- To delete one: click the trash icon next to the entry
- To delete all: click Clear all at the bottom
You can also manage memories from within a conversation:
- "What do you remember about me?" — ChatGPT lists its current memories
- "Forget that I am a teacher" — removes a specific memory
- "Clear all your memories about me" — wipes everything
How to Save Specific Facts
Explicit saves work better than hoping ChatGPT infers what to keep:
- "Remember that I work in a React/Node.js stack"
- "Remember that my client's company is called Acme Corp and they are in the healthcare sector"
- "Please remember I prefer responses in British English"
- "Note that I want you to always suggest tests when reviewing code"
ChatGPT confirms each save with a short message: "Got it, I'll remember that you prefer British English."
Temporary Chat: When to Turn Memory Off for One Session
Temporary Chat is a privacy mode that disables memory and history for a single conversation.
To start one: click the pencil/new chat icon → look for Temporary chat at the top, or click your avatar → Temporary chat.
Use it for:
- Sensitive topics you do not want associated with your account
- One-off tasks where your usual preferences do not apply
- Testing prompts without polluting your memory with test data
- Sharing your screen with others without exposing your history
Nothing from a Temporary Chat is saved, even if memory is globally enabled.
Common Mistakes to Avoid
- Not checking what has been saved — Memory can accumulate inaccurate facts over time. Check Manage memories monthly and delete anything outdated.
- Expecting memory to replace conversation context — Memory stores facts, not full conversations. If you want ChatGPT to continue a specific project, paste the relevant context at the start of the new chat rather than relying on memory alone.
- Relying on memory for sensitive project details — Do not rely on memory to store confidential client information or proprietary data. Memory storage goes through OpenAI's systems. Keep sensitive context in your own notes.
- Confusing memory with conversation history — History shows past chat logs. Memory stores extracted facts. Both live in Settings but serve different purposes. Deleting history does not delete memories, and vice versa.
- Using memory-heavy chats on shared accounts — Memory is per account. If multiple people use the same ChatGPT account, memories from one person's preferences will affect another person's sessions. Use separate accounts for this reason.
Memory vs. Custom Instructions: Which to Use
Both give ChatGPT persistent context, but they work differently:
| | Memory | Custom Instructions | |---|---|---| | Set by | ChatGPT (automatic) or you (explicit) | You (manual) | | Updated | Dynamically as you chat | Only when you edit settings | | Scope | Facts and learned preferences | Format and behavior rules | | Best for | Project context, personal facts | Tone, format, language style |
Use both together for the best experience. Custom Instructions for how you want responses formatted; Memory for what you are working on and who you are.