How ChatGPT Memory Works

ChatGPT’s current Memory experience is broader than a single list of saved facts. Depending on availability and settings, Memory may use saved memories, relevant information from past chats, custom instructions, Library files, or connected-app content. OpenAI’s Help Center notes that settings differ by plan, region, platform, and workspace. Some accounts may see controls such as Reference saved memories, Reference chat history, a memory summary, Manage, or a newer improved-memory experience. Saved memories are stored separately from chat history. They can include details the user asks ChatGPT to remember or information ChatGPT saves as useful context when that behavior is available. Reference chat history uses relevant past conversations to personalize future replies; the information derived from that history may change as ChatGPT updates what is useful to remember. A memory summary is a high-level view and may not show every source or detail used. Controls have different effects. Turning Memory off does not delete past chats. Turning Reference chat history off schedules information remembered from past chats for deletion within 30 days, while the original conversations remain unless deleted. Deleting a chat alone does not necessarily delete a separate saved memory. Removing information fully may require deleting the memory and its source chat, files, or connected-app source. Deletion and updates may take time to propagate. Temporary Chat does not create or update memories, though settings determine whether it uses existing personalization at the start. Users can review or correct available memory controls and should avoid entering sensitive information they do not want used for personalization. Workspace settings and data controls also matter; always check the current official Help Center for account-specific behavior.