ChatGPT Memory, Custom Instructions, and Temporary Chat Explained

You mention once that you manage a five-person sales team, and weeks later ChatGPT tailors an answer to "your team" without being told again. That is helpful when you want it and unsettling when you did not expect it. ChatGPT has three separate features that shape what it knows about you across conversations: memory, custom instructions, and temporary chat. They work differently, and the controls for each are in different places. The details below come from OpenAI's help pages as of October 2026.

One note before starting: OpenAI says memory features and controls can vary by plan, region, platform, and workspace settings, so a control described here may not appear in every account.

1. Memory: what ChatGPT can carry forward

OpenAI's Memory FAQ says that when memory is on, ChatGPT can remember relevant preferences and details. Depending on your plan and region, the sources it can draw on include:

• past chats;
• saved memories;
• custom instructions;
• files in your Library;
• content from connected apps, such as email.

It does not keep every detail of every conversation. OpenAI says ChatGPT decides which available information is relevant to a response, and that it looks for this context only when it is likely to improve an answer, not for every message.

There are two kinds of remembered information:

• Saved memories are details you explicitly ask ChatGPT to remember, or that it saves as useful context. OpenAI says they are stored separately from chat history.
• Reference chat history lets ChatGPT use relevant information from past conversations. OpenAI notes there is no separate storage limit for what it can reference this way.

2. How to see and change what it remembers

The controls are under Settings > Personalization > Memory. OpenAI describes a memory summary, a high-level view of information ChatGPT may use, and notes that it does not necessarily list every detail or source. You can also simply ask ChatGPT what it remembers about you.

When an answer uses your personal context, a Sources section may appear below it, showing a past chat, saved memory, instruction, file or email that contributed. OpenAI says these sources may not show every factor that shaped a response.

3. Deleting is more involved than it looks

This is the most important part for anyone who has shared work details. According to OpenAI:

• Deleting a chat does not necessarily delete a saved memory created from that chat. To fully remove something, delete the saved memory and the chat where you first shared it, and remove it from any files or connected apps where it also appears.
• "Don't mention this again" reduces future references but does not delete the underlying source.
• Turning off memory does not delete past chats. If you turn memory back on, ChatGPT may create new remembered information from chats still in your history.
• Turning off Reference chat history schedules information remembered from past chats for deletion within 30 days, while the chats themselves stay unless you delete them.
• OpenAI may keep logs of deleted saved memories for up to 30 days for safety and debugging.

4. Custom instructions: standing guidance you write

Custom instructions are different from memory. OpenAI's FAQ describes them as direct guidance you provide about what ChatGPT should know and how it should respond, while memory develops across conversations. The custom instructions help article adds:

• They apply to all chats, including existing conversations, as soon as you update them.
• Free and Go users can save up to 1,500 characters; Plus, Pro, Business, Enterprise and Education users up to 5,000.
• They are not shown to people who view a shared link.
• OpenAI says information from custom instructions is also used to improve model performance, and points to its data settings for opting out of content being used that way.

A good use at work: describe your role, preferred format, and tone ("Audience: nontechnical managers. Keep answers under 200 words."). A poor use: pasting client names, internal figures, or anything you would not want stored with your account.

5. Temporary chat: a conversation that stays out of history

OpenAI's Temporary Chat FAQ describes a temporary chat as one that stays out of your chat history and is not used to improve OpenAI models while it remains temporary. Before you send the first message, you choose:

• Unpersonalized: does not use memory, custom instructions, or plugins.
• Personalized: can use your existing memories and instructions.

Either way, a temporary chat does not create or update memories. You cannot change the choice after the conversation starts.

Two limits matter. OpenAI may keep a copy of a temporary chat for up to 30 days for safety purposes. And if you save a temporary chat, it becomes a regular chat that follows your normal personalization and model-improvement settings.

6. Which to use for work tasks

If you use a company workspace, your administrator may control some of these settings; OpenAI says workspace owners and admins can manage supported memory settings.

7. Key takeaways

• Memory can draw on past chats, saved memories, custom instructions, files, and connected apps, depending on plan.
• Deleting a chat does not necessarily delete a saved memory created from it; remove both.
• Turning off memory does not delete past chats.
• Custom instructions apply to all chats and are limited to 1,500 or 5,000 characters depending on plan.
• Temporary chats do not create memories, but a copy may be kept for up to 30 days for safety.