ChatGPT can remember useful details such as your food preferences, preferred writing style, or an ongoing goal, allowing later answers to feel more relevant. That convenience also creates an important question: exactly what does ChatGPT remember, and how can you correct or remove it?
Memory is not the same as chat history, Custom Instructions, or model training. Deleting a conversation may not remove a saved memory created from it, while turning memory off does not automatically erase information already remembered. Here is how the controls differ.
What is ChatGPT Memory?
OpenAI describes Memory as a personalization feature that uses useful context from chats, files and connected apps so you do not have to repeat yourself. Its availability and exact controls can vary by plan, account, region and rollout.
Depending on the memory experience available to your account, Settings may show a memory summary or controls for two related sources:
- Saved memories: Details ChatGPT saves for future conversations, including information you explicitly ask it to remember.
- Reference chat history: Useful context drawn from previous conversations rather than a short list of individually saved facts.
OpenAI has been updating the memory interface, so labels may differ slightly between web and mobile. The controls are generally found under Settings > Personalization > Memory.
Memory, chat history and Custom Instructions compared
| Feature | Purpose | Best used for | How to control it |
|---|---|---|---|
| Memory | Personalizes future replies using useful information about you | Preferences, recurring goals and relevant context | Settings > Personalization > Memory |
| Chat history | Stores conversations you can reopen or search | Returning to earlier discussions | Archive or delete individual chats under Data Controls |
| Custom Instructions | Provides explicit standing directions | Response style, units, role and formatting preferences | Settings > Personalization |
| Improve the model for everyone | Controls whether eligible consumer content may be used to improve models | Data-sharing preference | Settings > Data Controls |
| Temporary Chat | Starts a conversation without using or creating memories | One-off or more private discussions | Choose Temporary Chat before starting |
Turning off model training does not turn off memory, and turning off memory does not necessarily opt your chats out of model improvement. They are separate settings.
How to see what ChatGPT remembers
You can ask ChatGPT, “What do you remember about me?” for a conversational summary. For the account-level controls, open your profile menu and go to Settings > Personalization > Memory. Depending on your account, select the memory summary or Manage memories.
Review the information periodically. Preferences change: a temporary diet, old location, or completed project can make later recommendations less relevant if it remains in memory.
How to correct a wrong memory
Tell ChatGPT directly what has changed, such as “Forget that I am training for a marathon” or “Update my preference: I now want temperatures in Celsius.” Then review the Memory settings to confirm the change.
For an important correction, editing the memory control is more dependable than assuming a single message permanently changed every relevant source. Information may also exist in an older chat, uploaded file, or Custom Instructions.
How to delete a saved memory completely
OpenAI says that turning saved memory off does not delete information already remembered. Deleting the original chat also does not automatically remove a saved memory created from it.
To remove a saved memory:
- Open ChatGPT Settings.
- Select Personalization and open Memory.
- Choose the memory-management option available to your account.
- Delete the individual memory, or use the available option to clear memories.
To fully remove information, OpenAI recommends deleting both the saved memory and the conversation where the information was originally shared. If the detail is also present in Custom Instructions, a file, or a connected app, review that source separately.
What happens when you turn Memory off?
Memory controls stop ChatGPT from using the disabled memory source for personalization. They do not necessarily delete existing memories or past chats. OpenAI also notes that if memory is turned on again later, information from conversations that remain in chat history may become relevant again under the available memory experience.
If you want to keep memory enabled generally but exclude one conversation, use Temporary Chat before sharing the information.
Does Temporary Chat use Memory?
No. OpenAI says Temporary Chats do not use existing memories and do not create new ones. They are not placed in normal chat history and are not used to improve models, although OpenAI may retain a copy for up to 30 days for safety purposes.
Files uploaded in Temporary Chat are not saved to the user’s account or Library. Temporary Chat still follows OpenAI’s usage policies and should not be treated as an anonymous service. Avoid entering passwords, banking credentials, private medical records or other unnecessary sensitive data.
Does deleting a chat delete its memory?
Not necessarily. Saved memories are stored separately from chat history. Deleting the chat removes the conversation from view and schedules it for deletion under OpenAI’s retention policy, but a saved memory originating from that chat can remain until you delete it through Memory controls.
The reverse is also true: deleting a saved memory does not rewrite old conversation text. To remove both, manage the memory and the related chat separately.
How Memory works inside Projects
ChatGPT Projects maintain context from the chats and files within a project. Project-only memory can provide a self-contained working space that stays focused on that project rather than drawing from unrelated conversations. This is useful for a long-running research, writing, or planning workflow.
Review a Project before deleting it: OpenAI says deleting a Project permanently deletes its files, chats, and instructions and cannot be undone.
Should you keep ChatGPT Memory enabled?
Memory is useful if you regularly ask for personalized recommendations, maintain long-running goals, or prefer consistent response settings. Turn it off—or use Temporary Chat—when the information should not influence future conversations.
A practical middle ground is to keep only durable, non-sensitive preferences in memory: language, vegetarian food preference, preferred units or the audience for your writing. Avoid treating ChatGPT Memory as a password manager, medical record or secure document vault.
The bottom line
ChatGPT Memory, chat history, Custom Instructions, and training controls are separate systems. Turning Memory off does not erase saved memories, and deleting a chat may not delete a memory created from it. Review Memory under Personalization, remove incorrect entries, and delete the original conversation as well when you want information fully removed.
Sources: OpenAI Memory FAQ, OpenAI saved-memory guide, OpenAI Data Controls FAQ, OpenAI Projects guide and OpenAI retention policy.
