ChatGPT memory: every question answered.

ChatGPT memory is the feature that lets ChatGPT carry context about you from one conversation to the next. It has two layers: saved memories, which are explicit facts ChatGPT stores about you (name, role, preferences) that you can view and edit, and reference chat history, an implicit layer that draws on patterns from your past chats. Both are controlled in Settings → Personalization → Memory, and verified behavior is described on OpenAI's Memory FAQ.

Saved memories are auditable: you can open Manage memories, see each stored fact, and edit or delete it. Reference chat history (launched April 2025) is not a visible list — it's an implicit recall of your earlier conversations that you can only toggle on or off. So the saved-memories list is observable while the chat-history layer is not. Turning off saved memories doesn't erase existing memories unless you delete them.

Memory is stored on OpenAI's servers and tied to your account, not locally on your device — so it follows your login across devices. Every message you send is processed on OpenAI's servers regardless of your memory settings; the memory controls govern what's retained and reused afterward, not whether data leaves your device. Confirm current storage and retention details on OpenAI's help center.

No. The base model isn't retrained per user. Personalization happens at inference time: relevant saved memories and signals from past chats are injected into the model's context window alongside your prompt, like a hidden system note, so the model can condition its reply on them. The underlying weights don't change for your account. Exact implementation details are proprietary to OpenAI.

Go to Settings → Personalization → Memory and toggle "Reference saved memories" and "Reference chat history" independently. Note that, per OpenAI, turning off saved memories also turns off chat history. Turning memory off stops ChatGPT from using stored context but doesn't auto-delete your saved memories. For a one-off session with no memory at all, use Temporary Chat.

In Settings → Personalization → Memory, open Manage memories to delete individual saved memories or clear them all. You can also ask ChatGPT in chat to forget something. A saved memory persists even if you delete the conversation it came from, so deleting a chat doesn't necessarily remove the memory — you may need to delete it from Manage memories separately. Retention timelines are on OpenAI's help center.

You can see the full saved-memories list in Manage memories, but not the reference-chat-history layer, which has no visible list. So you can fully audit explicit memories but only infer the implicit signals drawn from past chats. To remove the implicit layer, turn off reference chat history and consider deleting old conversations.

For consumer accounts, OpenAI may use your conversations — including memory content — to improve its models by default. You can opt out in Settings → Data Controls by turning off "Improve the model for everyone." Business, Team, Enterprise and API data are not used to train models by default and are handled under different terms. Verify the current policy on OpenAI's Data Controls FAQ.

No. Temporary Chat doesn't appear in your history, doesn't read or write memories, and isn't used to train models. OpenAI says it keeps a copy for a short period (around 30 days) for safety and abuse monitoring, then deletes it. Use Temporary Chat whenever you want a private, memory-free conversation.

Not necessarily. A fact written to saved memories persists in the saved-memories list even if you delete the conversation it originated from, so you may need to delete the memory separately in Manage memories. Deleting chats does reduce what the reference-chat-history layer can draw on. Verify current behavior on OpenAI's help center.

Saved memory has a finite capacity. When it fills up, ChatGPT may show a "memory full" state and stop adding new saved memories until you delete some. OpenAI doesn't publish a precise size figure that we'd cite here. If memory is full, open Manage memories and remove outdated or unneeded entries to free space.

Common causes: memory is turned off in Settings → Personalization → Memory; you're in a Temporary Chat (which never uses memory); your saved memory is full; or the detail was never durable enough to be saved. Check that both "Reference saved memories" and "Reference chat history" are on, confirm you're not in Temporary Chat, and free space in Manage memories if it's full.

It can. Memory accumulates personal details over time, which is convenient but means sensitive information can persist across chats and resurface in unexpected contexts, and it's retained on OpenAI's servers. Reduce the risk by reviewing saved memories regularly, keeping financial, medical or credential data out of memory, using Temporary Chat for sensitive topics, and setting your data controls deliberately. Verify specifics on OpenAI's help center.

Conceptually, the system selects relevant saved memories and signals from your past chats and injects them into the model's context window alongside your new prompt, similar to a system note. The model then reads that added context and answers as if it remembers you. No per-user retraining occurs; personalization is an inference-time mechanism. The exact selection and injection logic is proprietary to OpenAI.

Yes. Business plans (Team, Enterprise) and the API generally offer stronger data protections, including not using business data to train models by default, plus admin-level controls. Memory availability and admin settings differ from consumer accounts. Check OpenAI's business documentation for current details on how memory and data controls apply to your plan.

Custom instructions are settings you write yourself that apply to every chat — for example, your preferred tone or background context. Memory is built up automatically by ChatGPT (saved memories) plus the implicit chat-history layer. Custom instructions are fully under your control and static until you change them; memory evolves as you chat and can be edited or deleted. Many users combine both.

No. Turning off "Reference saved memories" stops ChatGPT from using your stored facts, but it doesn't automatically delete them — they remain in Manage memories until you remove them. If you want the data gone, delete the memories individually or clear them all. Per OpenAI, turning off saved memories also turns off reference chat history.

OpenAI provides a data export option in Settings → Data Controls that lets you request a copy of your account data. The saved-memories list is also visible directly in Manage memories, where you can read each entry. For the exact contents and format of an export, and how the reference-chat-history layer is represented, check OpenAI's Data Controls FAQ, as these details can change.

Yes. Because memory is tied to your account on OpenAI's servers rather than to a device, it follows you when you sign in on the web, the mobile app or other clients. There's no separate per-device memory. Signing out and into a different account gives you that account's memory instead.

Decide deliberately what you want personalized, keep sensitive data (financial, medical, credentials) out of memory, review the saved-memories list periodically, use Temporary Chat for one-off or private tasks, and set your data controls — including the training opt-out — the way you want. If you use several AI tools, remember each one stores context separately, so spreading usage limits how much any single provider accumulates.