ChatGPT Memory Settings: Complete Control Guide 2026
ChatGPT's memory feature saves facts about you across conversations. It is useful when it works well and frustrating when it remembers the wrong things or refuses to forget. This guide covers every way to control what ChatGPT remembers, how to turn memory off when you need privacy, and when to stop using memory entirely.
How ChatGPT Memory Works
ChatGPT memory is an automatic cross-conversation storage system. As you chat, the model extracts bullet-point facts and saves them to a memory store. Future conversations load these facts and use them to personalize responses.
Examples of what gets saved:
- Your name, role, and preferences.
- Your tech stack or tools.
- Projects you mention repeatedly.
- Tone preferences like "User prefers concise answers."
What does not get saved:
- Full conversation transcripts.
- Detailed project state or architecture decisions.
- Temporary information from one-off chats.
Memory is optimized for personal facts, not project continuity. If you need the latter, read Why AI Forgets Your Project Context Between Sessions.
Where to Find Memory Settings
All memory controls live in the same place:
- Open ChatGPT.
- Click your profile icon (bottom-left corner).
- Go to Settings.
- Navigate to Personalization.
You will see two options:
- Memory — the main toggle. Controls whether ChatGPT saves new memories going forward.
- Manage Memory — view and delete individual memories.
How to Turn Memory On
If memory is off and you want to enable it:
- Go to Settings → Personalization → Memory.
- Toggle the switch to On.
From this point forward, ChatGPT will save facts from your conversations. Existing memories (if you had any before turning it off) remain intact.
How to Turn Memory Off
If you want ChatGPT to stop saving new memories:
- Go to Settings → Personalization → Memory.
- Toggle the switch to Off.
What this does:
- ChatGPT will not save new memories from future conversations.
- Existing saved memories remain in storage unless you delete them manually.
- Each conversation still has its own context window — turning off memory only affects cross-conversation storage.
What this does not do:
- It does not delete existing memories.
- It does not prevent ChatGPT from using context within a single conversation.
How to View Your Saved Memories
To see what ChatGPT has saved about you:
- Go to Settings → Personalization → Manage Memory.
- Scroll through the list.
Each memory appears as a bullet-point fact with a small delete button (trash icon) next to it. The list shows everything ChatGPT has stored — your preferences, your projects, and any other facts the model deemed worth saving.
How to Delete a Specific Memory
If ChatGPT saved something incorrect or something you no longer want stored:
- Go to Settings → Personalization → Manage Memory.
- Find the memory you want to remove.
- Click the trash icon next to it.
- Confirm deletion.
The memory is removed immediately. ChatGPT will no longer use that fact in future conversations. If you mention the same topic again, ChatGPT may save a new memory about it unless memory is turned off.
How to Clear All Memories at Once
If you want to wipe the slate clean:
- Go to Settings → Personalization → Manage Memory.
- Scroll to the bottom.
- Click Clear All Memories.
- Confirm.
This deletes every saved memory. ChatGPT will start from zero the next time you chat. The memory toggle remains in whatever state you left it — if memory is on, ChatGPT will start saving new memories immediately.
How to Use Temporary Chats (Per-Conversation Memory Control)
If you want to have a single conversation without contributing to memory — for example, a sensitive or one-off topic — use Temporary Chat.
How to start a temporary chat:
- Open ChatGPT.
- Click the model picker or chat options at the top.
- Select Temporary Chat.
Temporary chats:
- Do not save to your conversation history.
- Do not contribute to saved memory, even if the global memory toggle is on.
- Are deleted when you close the tab.
If you need to keep any part of a temporary chat, press Ctrl+S (Cmd+S on Mac) to save the page as HTML before closing. More on this workflow: How to Back Up All Your AI Conversations in One Place.
When to Turn Memory Off Entirely
You should consider turning memory off if:
You work with sensitive information. Client data, legal research, medical notes, or product strategy should not accumulate in vendor-controlled storage. Turn memory off and use temporary chats for sensitive work.
ChatGPT keeps saving incorrect facts. If you find yourself deleting and correcting memories multiple times a week, the feature is causing more friction than it solves.
You switch between multiple AI platforms. ChatGPT memory only works in ChatGPT. If you also use Claude or Gemini, the memory fragmentation across platforms makes vendor memory less useful. Consider using a portable memory layer instead: Give ChatGPT, Claude, and Gemini Persistent Memory Across Every Chat.
You share an account. If multiple people use the same ChatGPT account, memory will mix facts from different users and produce inconsistent behavior. Turn it off and manage context manually.
Common Memory Settings Problems and Fixes
Problem: Memory toggle is on, but ChatGPT is not saving new memories.
Possible causes:
- The memory store is full. ChatGPT has a storage cap. When it fills up, new memories stop saving or old ones get dropped.
- You are using a temporary chat, which bypasses memory even if the global toggle is on.
Fix: Go to Manage Memory and delete old or irrelevant entries to free up space. Confirm you are in a normal chat, not a temporary one.
Problem: I deleted a memory, but ChatGPT still acts on it.
Cause: The context window of the current conversation still contains the deleted fact. Deleting a memory removes it from cross-conversation storage, but it does not erase it from the conversation you are already in.
Fix: Start a new chat. The deleted memory will not appear in future conversations.
Problem: Memory is off, but ChatGPT still seems to remember me.
Cause: You are conflating memory with the context window. Even with memory off, ChatGPT can see everything you said earlier in the current conversation. Turning off memory only affects cross-conversation storage.
Fix: If you want full privacy, use a temporary chat in an incognito browser window and do not log in.
Problem: ChatGPT saved a wrong fact and keeps using it.
Fix: Go to Manage Memory, find the incorrect entry, delete it. Then tell ChatGPT the correct information in a new chat. If memory is on, ChatGPT will save the corrected fact.
Memory vs Context Window: Know the Difference
This is the most common source of confusion. ChatGPT has two separate systems:
- Context window — the text ChatGPT can see within a single conversation. This resets to zero when you start a new chat.
- Saved memory — bullet-point facts stored across conversations. This persists between chats if the memory toggle is on.
Turning off memory does not shrink the context window. It only stops ChatGPT from saving new cross-conversation facts.
If ChatGPT forgets something you said earlier in the same conversation, that is a context window problem, not a memory problem. Covered in ChatGPT Loses Context Mid-Chat? Here's The Real Fix.
Controlling Memory on Mobile
The memory settings are identical on mobile (iOS and Android):
- Open the ChatGPT app.
- Tap your profile icon.
- Go to Settings → Personalization.
- Toggle Memory or tap Manage Memory.
Temporary chats work the same way: tap the model picker at the top of a new chat and select Temporary Chat.
Privacy and Where Memory Is Stored
ChatGPT memory is stored on OpenAI's servers. It is tied to your account and syncs across devices. This means:
- Your memories are visible to OpenAI.
- They may be subject to OpenAI's data retention policies.
- They can appear in account exports, security incidents, or legal requests.
For work that requires true data sovereignty, turn memory off and keep context in a local-first memory layer. Details: Private AI Memory: Using Incognito Mode With Full Data Sovereignty.
Exporting Your Memories
ChatGPT does not have a dedicated memory export button, but your saved memories are included in the full account data export.
To export:
- Go to Settings → Data Controls → Export Data.
- OpenAI emails you a .zip file within a few hours.
- Inside the archive, look for the file containing memory entries (typically included in the user data JSON).
For more on what the export contains: What to Do With Your ChatGPT Data Export File in 2026.
When to Stop Using Memory Entirely
If any of these apply, you should turn memory off and manage context manually:
- You delete or correct memories more than once a week.
- You work on multiple unrelated projects and the memory mixes them.
- You need your context to move across ChatGPT, Claude, Gemini, or other platforms.
- Your work involves sensitive information that should not live on a vendor's server.
The alternative is a portable memory workflow: save conversations, distill them into memory documents, and paste the relevant context into each new chat. This works across every AI platform and gives you full control. Setup guide: How to Build a Personal Knowledge Base From AI Chats.
Real-World Memory Management Scenarios
Understanding the controls is one thing. Knowing when and how to use them in practice is another. Here are common scenarios and how to handle them with ChatGPT memory settings:
Scenario 1: Starting a new unrelated project. You have been using ChatGPT for your web development work, and now you are starting a hardware project. The old memories about your stack, your preferences, and your past projects are irrelevant and may confuse the model. Solution: Go to Manage Memory and delete memories tied to the old project. Keep general preferences like tone and format. ChatGPT will save new memories about the hardware project as you work.
Scenario 2: Interviewing job candidates. You need to use ChatGPT to help draft interview questions, but you do not want the names, companies, or details of candidates saved to memory. Solution: Use a temporary chat for each interview-related session. This prevents any candidate information from entering your permanent memory store.
Scenario 3: Collaborating with a client. A client asks you to use ChatGPT to draft a proposal. The proposal contains confidential product details and pricing that should not persist in your account. Solution: Turn memory off before starting the chat, or use a temporary chat. After the work is done, confirm that no memory entries were created by checking Manage Memory.
Scenario 4: ChatGPT remembers an outdated preference. Six months ago you told ChatGPT you preferred Python. You have since switched your entire workflow to TypeScript, but ChatGPT still suggests Python solutions. Solution: Go to Manage Memory, find the outdated "User prefers Python" entry, delete it. In the next chat, mention your current preference. ChatGPT will save "User prefers TypeScript" if memory is on.
Scenario 5: Sharing your ChatGPT account with a team. Multiple people use the same organizational ChatGPT account. Memory mixes facts from different users and produces inconsistent behavior. Solution: Turn memory off entirely for shared accounts. Each user should manually provide context at the start of their chats, or use a memory layer that each person controls separately.
These scenarios show that memory settings are not set-it-and-forget-it. The right configuration depends on the task, the sensitivity of the work, and whether you are working solo or collaboratively.
Summary of All Memory Controls
| Action | Path | Effect |
|---|---|---|
| Turn memory on | Settings → Personalization → Memory (toggle on) | ChatGPT saves new memories from future chats |
| Turn memory off | Settings → Personalization → Memory (toggle off) | ChatGPT stops saving new memories; existing memories remain |
| View saved memories | Settings → Personalization → Manage Memory | See all stored facts |
| Delete a specific memory | Manage Memory → trash icon next to entry | Remove one fact; others remain |
| Delete all memories | Manage Memory → Clear All Memories | Wipe all stored facts |
| Single private chat | Start Temporary Chat | One conversation with no memory contribution and no history save |
| Full privacy mode | Incognito window + temporary chat + memory off | Maximum privacy; nothing persists |
The Bottom Line
ChatGPT memory is a convenience feature. When it works, it saves you from repeating personal facts. When it does not, the controls are straightforward: turn it off, delete what you do not want, or use temporary chats for sensitive work.
If you need reliable project continuity across platforms, vendor memory is not the answer. A portable memory layer that you control gives you the continuity without the lock-in.
For tools that make portable memory practical: MindLock.