ChatGPT Lets Users Personalize Temporary Chats Without Saving Them to History
OpenAI updated ChatGPT’s temporary chat feature on August 27, 2026, giving users more control over how much personalization a private conversation can use. Previously an all-or-nothing privacy mode, temporary chat now lets users choose whether to bring in memory, custom instructions, and plugins, or save a conversation afterward so it isn’t lost when the session ends.
Details
- Personalized mode: a temporary chat can now reference existing memories, custom instructions, and plugins to tailor its answers, without creating new memories or appearing in chat history unless the user later saves it
- Non-personalized mode: remains the default behavior, excluding memory and plugin access entirely, as temporary chat has worked until now
- Locked at start: the personalization choice is set only when a temporary chat begins and cannot be changed mid-conversation
- Save option: users can convert a temporary chat into a permanent one, at which point it becomes subject to normal account-level personalization and model-improvement settings
- Rollout: available now across ChatGPT’s supported plans
What happened next
The change addresses a common complaint that temporary chat’s privacy benefits came at the cost of usefulness, since it previously ignored a user’s saved context entirely. By splitting personalization from history-saving, OpenAI lets privacy-conscious users keep the option to use their existing memory and instructions on a one-off basis, while still defaulting to a fully clean slate unless they opt in.