To start a Temporary Chat in ChatGPT, begin a new chat and look for Temporary or the temporary-chat control in the new-chat area; turn it on before you send your question. It is meant for a one-off conversation: it does not appear in your normal chat history and does not create or use memories. That does not mean it is a vault for sensitive information. Account-level data controls are separate, and OpenAI may retain a copy for a limited period for safety purposes. If the labels on your screen differ, use OpenAI’s current Temporary Chat FAQ.
Know what Temporary Chat is for
Temporary Chat is useful when a question does not belong beside your ongoing work. Maybe you want help comparing two versions of a short note, brainstorming a party menu, or turning a rough idea into a few neutral questions. You want the help, but you do not need the conversation waiting in your sidebar later or informing future personalization.
Think of it as a clean, short-lived workspace—not a different kind of assistant and not a guarantee that every system connected to your life forgets the exchange. The chat can still answer based on what you type in that conversation. It can also still make mistakes. For a decision that matters, ask for sources and check the original information yourself.
Use an ordinary chat when the conversation is part of a project you expect to revisit. For example, a continuing work plan is easier to manage in ChatGPT Projects, where you intentionally keep related context together. Temporary Chat is the better fit when keeping that context would add clutter rather than value.
Start a Temporary Chat
The exact placement can change as ChatGPT’s interface changes, but the routine is simple:
- Open ChatGPT and start a new conversation.
- In the new-chat controls, select the Temporary option or temporary-chat icon before entering your prompt.
- Check that the conversation is marked as temporary, then ask your question.
- When you are done, close the conversation and avoid treating it as a note you can return to from history.
If you cannot find the option, do not guess based on an old screenshot. Check the official FAQ linked above and make sure you are signed into the account you intend to use. Feature placement and availability can vary by app, account, and product update.
A good first prompt is small and specific: “Help me make this invitation sound warmer in two sentences. Do not add facts.” A narrow request gives you an answer that is easier to inspect. It also keeps you from pasting an entire private thread when a summarized version would do.
History, memory, and data controls are different things
These three ideas often get blended together, which leads to bad assumptions.
Chat history is the list of saved conversations you normally see and reopen. A Temporary Chat does not show up there.
Memory is information ChatGPT may use to make future conversations more tailored. According to OpenAI’s Temporary Chat guidance, a Temporary Chat does not create memories or use existing memories.
Data controls are account settings that govern choices such as whether conversations can be used to improve models. They apply at the account level; starting a Temporary Chat is not a substitute for reviewing them. OpenAI explains the current choices in its Data Controls FAQ.
There is another important boundary. OpenAI says it may keep Temporary Chats for up to 30 days and may review them to monitor abuse. So “not in my history” is a useful promise about the ChatGPT experience, but it should not be translated into “nobody can ever access this” or “nothing is retained anywhere.” If you need to handle protected, confidential, regulated, or highly sensitive material, follow your organization’s rules and use an approved tool instead.
Use a quick privacy filter before you paste
Temporary Chat can reduce clutter, but it is not a reason to paste more than necessary. Before sending a prompt, remove or generalize details ChatGPT does not need:
- Names, home addresses, phone numbers, and email addresses
- Passwords, recovery codes, account numbers, and payment-card details
- Medical records, identity documents, and private messages from other people
- Confidential client, employer, school, or legal information
Try a substitute description first. Instead of pasting a customer’s message, write: “A customer is upset about a delayed order. Help me draft a calm reply that acknowledges the delay without promising a refund.” You retain the decision-making and private context; the assistant gets enough information to help with wording.
The same habit matters in voice conversations. Our guide to using ChatGPT Voice for an action plan shows how to turn an idea into a checklist without dictating private details. A temporary setting is helpful, but minimizing sensitive input is the stronger everyday practice.
Review your account controls separately
If your real question is “How is my account configured?”, take a few minutes to inspect Settings and the data controls available to you. Read the description beside each option rather than assuming a switch covers history, memory, model improvement, and account security all at once. OpenAI’s help page is the source to trust when its product wording changes.
You can also review and manage memories independently in the relevant personalization settings. If you are concerned that someone else may be using your account, Temporary Chat is not the remedy. Go to Settings > Security > Active sessions and follow our guide to reviewing unfamiliar ChatGPT sessions. Sign out sessions you do not recognize, then protect the account itself.
Get useful answers without overtrusting them
Temporary Chat changes where a conversation appears and how it relates to memory; it does not make every response correct. Read the output before sending it, especially if it contains facts, calculations, dates, links, or advice that affects money, health, legal matters, safety, or another person.
For a one-off question, this little closing prompt helps: “List any assumptions you made and tell me which facts I should verify.” It turns a polished-looking answer into something you can check. Then verify important details with the original source, not with confidence or wording alone.
Official sources
- OpenAI Help Center: Temporary Chat FAQ
- OpenAI Help Center: Data Controls FAQ