ChatGPT’s Temporary Chat is no longer necessarily a context-free conversation. OpenAI now documents two modes that users choose before the session begins: Personalized, which can use existing memories, custom instructions and plugins, and Unpersonalized, which does not use those personalization sources.
The updated official Temporary Chat documentation also changes another important assumption about the feature: a temporary conversation does not have to remain temporary forever. Users can save it to chat history, converting it into a regular conversation. If the session contains uploaded files, eligible files may also be saved to Library when the chat is saved.
The result is a more flexible model than the original idea of an isolated blank-slate session. “Temporary” now primarily describes how the conversation is stored and handled while it remains temporary; it does not automatically mean “without existing personal context.”
Temporary Chat now starts with a personalization choice
OpenAI says the choice is made before the first message. After selecting Temporary, the user chooses Personalized or Unpersonalized and then begins the conversation.
A Personalized Temporary Chat can use existing memories, custom instructions and plugins to tailor its responses. An Unpersonalized Temporary Chat does not use those sources.
The choice cannot be changed after the conversation starts. OpenAI also says account and workspace restrictions take priority over the personalization choice for an individual Temporary Chat.
Personalized does not mean the temporary session creates new memories
The distinction between using memory and creating memory is crucial. OpenAI says a Personalized Temporary Chat can reference memories that already exist, but neither personalization mode creates or updates memories while the conversation remains temporary.
That means a user can receive responses informed by existing remembered context without the temporary conversation itself becoming a new source of memory at that stage.
This is one of the most important nuances in the updated model. Temporary Chat is not necessarily memory-free, but it remains memory-non-writing while temporary.
Unpersonalized mode is the explicit blanker-slate option
Users who do not want existing memory, custom instructions or plugins involved in the session can select Unpersonalized before sending the first message.
That is now the clearer choice for someone who wants to reduce personalization context for a particular conversation. OpenAI’s Memory documentation likewise tells users to choose a non-personalized Temporary Chat when they do not want the conversation to use memories for personalization.
There is still a safety qualification. OpenAI says ChatGPT may use limited information from previous conversations for safety and security purposes in rare high-risk situations. Turning personalization off does not disable those safety systems.
Plugins are available only in Personalized Temporary Chats
The updated documentation explicitly includes plugins in the personalization choice. Personalized Temporary Chats can use plugins that are otherwise available to the account or workspace, while Unpersonalized Temporary Chats cannot.
This makes Temporary Chat more capable than a simple isolated text session. A user can choose a temporary conversation that remains outside normal history while still benefiting from existing personalization and connected capabilities.
At the same time, plugin use introduces an external-data boundary. OpenAI notes that when a GPT uses actions, data sent to a third party through those actions is governed by the recipient’s privacy policy, and the third party may retain that data for longer than OpenAI’s temporary-chat retention period.
A Temporary Chat can now be saved after the fact
OpenAI says users can save a Temporary Chat to their chat history if the conversation becomes useful enough to keep. Saving converts it into a regular chat, including when the conversation originally began in Unpersonalized mode.
From that point forward, the conversation follows the account’s normal personalization and model-improvement settings. If Memory is enabled, OpenAI says the saved conversation may subsequently be used to personalize future responses.
This creates a meaningful change in user workflow. A person no longer has to know at the beginning whether a conversation deserves permanent history. They can begin temporarily and decide later that the work is worth preserving.
Saving changes the data rules that apply to the conversation
While a conversation remains temporary, OpenAI says it is not used to improve its models, regardless of whether the user selected Personalized or Unpersonalized mode. OpenAI may retain a copy for up to 30 days for safety purposes.
If the user saves the conversation, it becomes a regular chat and the account’s model-improvement setting applies. The conversation also moves to the retention rules for saved chats.
That transition is important for privacy expectations. Saving is not merely pinning a temporary session to the sidebar; it changes the conversation’s status and the settings that govern it.
Eligible uploaded files can move into Library when the chat is saved
The updated Temporary Chat documentation also connects the feature with ChatGPT Library. OpenAI says that if a temporary conversation contains uploaded files, eligible files may be saved to Library when the chat itself is saved, subject to availability and storage limits.
OpenAI’s Library documentation confirms the same lifecycle. Files uploaded in Temporary Chat are not saved to Library while the conversation remains temporary. If the user saves the chat to history, eligible files may then be saved.
This means the save action can preserve more than conversational text. It can turn a previously temporary working session into a reusable combination of chat history and eligible file assets.
Temporary files and Library files follow different lifecycles
The distinction matters because Library files are managed separately from chat history. OpenAI’s retention documentation says files saved to Library can remain available independently of the conversation that used them.
Deleting a chat therefore does not automatically delete a file that remains saved in Library. A user who wants to remove both needs to manage the saved Library copy as well.
For Enterprise, Edu and Healthcare workspaces, Library files follow applicable workspace retention policies.
“Temporary” now describes persistence more than isolation
The updated architecture changes the most intuitive interpretation of the feature. A Temporary Chat can now be personalized with information that existed before the conversation, can use plugins, and can later be promoted into normal history.
What remains temporary is the session’s status while the user leaves it unsaved: it stays out of normal chat history, does not create or update memories, and is not used to improve OpenAI’s models while temporary.
That is a different concept from a fully isolated session that knows nothing about the user.
The distinction matters for privacy-sensitive workflows
Users may historically have selected Temporary Chat because they wanted a conversation outside normal history. That still makes sense, but history and personalization are now clearly separate choices.
A Personalized Temporary Chat can stay outside history while using existing memories and instructions. An Unpersonalized Temporary Chat can stay outside history without using those personalization sources.
For privacy-sensitive or controlled testing workflows, users therefore need to pay attention to both dimensions: whether the conversation is temporary and whether it is personalized.
It also changes how developers and researchers should test ChatGPT behavior
Temporary Chat is often useful when comparing prompts or trying to reduce the influence of prior conversation history. Under the documented behavior, selecting Temporary alone is no longer sufficient to establish that memory, custom instructions and plugins are absent.
For a cleaner personalization test, the relevant mode is Unpersonalized Temporary Chat. Otherwise, existing account context may still influence the response even though the conversation itself is not being written into memory.
This distinction is especially relevant for researchers documenting model behavior, support teams reproducing user experiences and publishers comparing responses across accounts.
Saving an Unpersonalized Temporary Chat changes what can happen later
OpenAI explicitly documents the edge case of saving a conversation that began without personalization. Once saved, it becomes a regular chat and follows the account’s normal personalization settings.
If Memory is active, the saved conversation may then become relevant to future personalization. In other words, Unpersonalized describes how the conversation operates while temporary; it does not permanently mark the conversation as excluded from memory if the user later saves it.
Account and workspace restrictions continue to apply throughout the transition.
Temporary Chat now has three distinct states to think about
Conceptually, the updated feature creates three useful states. A Personalized Temporary Chat can read existing personalization context without creating new memory. An Unpersonalized Temporary Chat operates without memory, custom instructions or plugins. A saved Temporary Chat stops being temporary and becomes a normal conversation governed by regular account settings.
Those states make the product more flexible, but they also make the word “temporary” less descriptive on its own. Users need to know whether the session is personalized and whether they later choose to save it.
The change fits ChatGPT’s broader move toward reusable personal context
The update arrives as ChatGPT increasingly treats past work, files and connected information as reusable context. NetContentSEO recently documented how ChatGPT’s unified search now spans chats, projects, images, documents, Library files and eligible connected Google Drive content.
Temporary Chat provides the opposite control surface within that broader architecture. Users can start outside normal history, decide how much existing personalization to bring into the session, and later choose whether the resulting work should join the persistent corpus.
That makes temporary and persistent work less like two completely separate modes and more like a lifecycle the user can control.
Temporary Chat is now a choice about context and persistence
OpenAI’s updated documentation makes one point especially clear: a Temporary Chat should no longer automatically be described as a blank slate.
Personalized mode can use existing memories, custom instructions and plugins. Unpersonalized mode removes those personalization sources. Neither mode creates new memories while the session remains temporary, and neither is used for model improvement during that temporary period. The user can then save the conversation, at which point it becomes a normal chat and eligible uploaded files may also move into Library.
The practical privacy question is therefore no longer simply “Is this a Temporary Chat?” It is also “Is this Temporary Chat personalized, and will I save it later?” Those choices now determine how much existing context enters the conversation and whether the session ultimately becomes part of the user’s persistent ChatGPT workspace.